TOPIC
Agentic Commerce
A subject the collection covers, read through 145 synthesis notes.
View as
Can cheap agent search actually harm shopping markets?
A 2026 market design paper challenges the assumption that near-zero search costs improve competition. It explores whether imperfect agent search, convergent descriptions, and easy entry can combine to leave markets worse off than before.
Can AI seller agents shift buyer preferences through description edits? Do shopping agents respond more strongly to price and rating cues? Can platform ranking and seller descriptions separately harm AI shoppers? Will brands succeed by building their own AI shopping agents? Do LLM agents achieve fair marketplace outcomes when given ideal search?
AI seller agents shifted market share by editing product descriptions alone Shopping agents shift choices far more than humans under identical manipulations Platform ranking and sellers' descriptions compound consumer harm more than doubling independently Consumers expect AI agents to reshape purchasing but infrastructure gaps remain LLM agents approach optimal welfare only under ideal search conditions
Can agent platforms profit more from search than consumers benefit from it? Does preference articulation become scarcer as search costs fall? What happens to market entry and product variety when generation costs drop? Does mixing agent types reduce unexplored product alternatives in markets? Does bilateral agent commerce expand the discoverable frame more than unilateral agent search? Why might agentic search become profitable for platforms before helping consumers? How can brands become the agent's default choice instead of competing for attention? What role would shortlist inclusion play if agents eliminate sponsored ad visibility? Can platforms profit from agentic search before consumers actually prefer it? How does MCP's zero-commission model compare to current OTA commission structures? What evidence shows communication costs currently limit consumer choices? Can agentic search become revenue-superior for platforms before benefiting consumers?See all 21 inquiring lines on this note →
Can Grice's maxims explain LLM advertising conflicts?
Does framing chatbot advertising violations through linguistic cooperation principles—Quality, Quantity, Relevance, Manner—help identify and regulate conflicts of interest between company incentives and user welfare?
Do LLMs prioritize company profits over user interests in ads? How do we learn to read AI-generated text critically? Does telling people an AI wrote something actually stop them from believing it? Do language models favor expensive sponsored products over cheaper alternatives? Does naming the platform change how AI agents judge sponsored listings?
Most LLMs favor company incentives over user welfare in sponsored recommendations AI discourse circulates without the cultural discount we apply to advertising AI disclosure raises scrutiny but leaves meaningful residual persuasive effect intact Current language models favor sponsored products over cheaper identical alternatives even at double the price Platform delegation reduces AI agent skepticism toward sponsored hotel listings
Why do standard web analytics fail to credit brand mentions in chat? Could disclosure requirements prevent agents from hiding advertising inside take rates? Could referral fee structures create conflicts between agent loyalty and customer interest? Do sponsored agents or advertising-funded assistants face hidden incentive conflicts? How does OpenAI prevent advertising from influencing ChatGPT answer quality? Should consumer law concepts like best interest duties be redesigned for AI agents? How should chatbots disclose sponsored recommendations to users? Which LLMs favor company incentives over user welfare in conflicts? Are these seven scenarios exhaustive for all advertising conflicts? How do people discount advertising claims differently than AI generated text? Does this sponsorship bias pattern appear in other domains like financial advice? Can companies be held liable for statements made by their deployed chatbots?See all 14 inquiring lines on this note →
Do users trust extroverted AI advisors despite worse advice?
When LLM financial advisors adopt different personalities, does user trust follow the quality of advice given, or do personality traits override information quality in shaping confidence?
Does chatbot personalization build trust or expose privacy risks? Can a seller infer what a buyer really wants to pay? Does personalized LLM advice actually help investors make better decisions? Does co-design participation hide misalignment in preference agents? Does personalization make large language models worse at their jobs?
Chatbot personalization increases trust and privacy risk simultaneously Sellers infer willingness to pay nearly perfectly from shopping dialogue alone Personalized LLM financial advice helps investors only when preference elicitation succeeds Co-design participation may mask agent misalignment through systematic overtrust Personalization consistently degrades LLM behavior across three key dimensions
What reputation signals do LLM assistants fail to register at all in their recommendations? Can advisor persona traits override the actual quality of financial recommendations? Why do users trust personalized financial advisors even when advice quality suffers? How can investors detect when AI personalization is degrading rather than improving advice? What distinguishes governance risks specific to LLM personas from general CUI ethics? Does repeated exposure to wrong LLM advice change confidence calibration over time? Do LLM confidence signals actually drive human overreliance on advice?
Can platform ranking and seller descriptions separately harm AI shoppers?
A simulation tested whether a platform's ranking algorithm and sellers' product descriptions might each independently exploit agent biases. Understanding this matters because it could reveal how uncoordinated actors produce coordinated-looking market harm.
Can AI seller agents shift buyer preferences through description edits? Do AI shopping agents show persistent position bias in text interfaces? Can cheap agent search actually harm shopping markets? Will brands succeed by building their own AI shopping agents? Do platforms and sellers exploit AI shoppers together more effectively?
AI seller agents shifted market share by editing product descriptions alone AI shopping agents show persistent position bias resistant to prompting and interface changes Cheap agent search can leave markets worse off despite lower costs Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Platform and seller exploitation combined harms consumers over twice as much.
Should retailers be able to refuse agents before any commercial relationship begins? Can sellers influence their visibility through caching layer mechanics? Can self-report feedback detect agent quality gaps in real markets? Can retailers use off-site agents to compete without losing customer loyalty? Why do shoppers trust retailer-built agents more than third-party AI agents? Why would a platform prefer to control basket assembly rather than just product search? Does mixing agent types reduce unexplored product alternatives in markets? How do existing consumer protection laws handle algorithmic agents as decision-makers? Can brands become preferred by AI agents without platform control? Can seller agents responding to AI shoppers actually shift market share between competitors? What prevents human-run markets from discovering valuable multi-party transaction combinations? What other agent beliefs might causally influence their pricing decisions?See all 62 inquiring lines on this note →
How do seller subsidies sort quality in search markets?
When sellers can pay to lower buyer inspection costs, do they sort by quality? This explores whether subsidies act as credible signals and how consumers should rationally order their search.
Does platform pricing of inspection tokens encourage excess search? Why does manual search fail as product choices become complex? Why do consumers delay adopting agentic search? Can an algorithm recommend products without matching an ideal mechanism? Does algorithmic recommendation lose its edge as sellers compete more?
Platform pricing of inspection tokens induces excess search and seller harm Agentic search escapes complexity collapse by prioritizing preference articulation Platforms want agentic search before consumers rationally should switch Algorithmic recommenders match optimal mechanism welfare using posted prices as cost signals Algorithmic recommendation advantage shrinks as seller competition intensifies
What role would shortlist inclusion play if agents eliminate sponsored ad visibility? How does cheap seller description generation reduce offer differentiation? Why do subsidy caps create pooling at the top instead of perfect separation? What happens when a platform prices inspection tokens in this subsidy model? How does subsidy signaling compare to auction-based ranking mechanisms? Can sellers credibly signal quality through inspection cost subsidies alone? How does pooling at subsidy caps drive excessive inspection relative to social optimum? When does marking up inspection access benefit platforms while harming seller surplus? Why do buyer-aligned objectives change how an algorithm weights seller costs?
Do language models favor expensive sponsored products over cheaper alternatives?
A new study tested whether AI models steer users toward pricier sponsored items even when identical cheaper options exist. This matters because it reveals potential conflicts of interest in AI shopping assistants as advertising enters their deployment.
Should AI shopping agents face fiduciary duties like human brokers? Can traditional governance tools actually control AI agents? Can AI seller agents shift buyer preferences through description edits? Why do consumers delay adopting agentic search? Do LLMs prioritize company profits over user interests in ads?
Owens and Law argue AI agents need fiduciary duties before markets harden Conventional governance tools fail for opaque, fast-moving AI agents AI seller agents shifted market share by editing product descriptions alone Platforms want agentic search before consumers rationally should switch Most LLMs favor company incentives over user welfare in sponsored recommendations
How often do AI assistants recommend the same brand across different conversations? How do AI assistants filter which retail links they recommend to users? Do AI recommendations influence user choice differently than human advice does? Why do companies disguise premium prices as discounts in billing? Why might eco-certification signals carry different weight in machine versus human hotel choice? What affiliate fee rates do different merchants currently offer for referrals? Do sponsored agents or advertising-funded assistants face hidden incentive conflicts? What usage patterns dominate free tier conversations before ads launch? What evidence shows communication costs currently limit consumer choices? How do advertising incentives change what AI models recommend to users? Do users apply the same discount to AI ads as human ads? Are these seven scenarios exhaustive for all advertising conflicts?See all 20 inquiring lines on this note →
Can a seller infer what a buyer really wants to pay?
Does a buyer's natural shopping dialogue reveal their willingness to pay to a seller who reads only the conversation? An experiment tests whether verbal behavior patterns leak financial information without explicit numbers.
Does chatbot personalization build trust or expose privacy risks? Does personalization make large language models worse at their jobs? Can better prompts stop preference leakage in AI shopping? Do agent representation failures or negotiation skills cause market shortfalls? Is Google's personalized pricing actually surveillance pricing?
Chatbot personalization increases trust and privacy risk simultaneously Personalization consistently degrades LLM behavior across three key dimensions Preference leakage in AI shopping arises from personalization itself, not prompt failures Representation failures, not negotiation skill, caused most market shortfall Google's personalized pricing is surveillance pricing with disguised premiums
How do different AI models weigh price signals versus product description text? What conflicts exist between vendor incentives and objective consumer research findings? Does stated trust in surveys predict actual delegated purchasing behavior? Will merchants deliberately exploit stable position biases to influence agent purchasing? Which LLM models consistently win better deals as both buyer and seller? Does visible transaction history reduce or increase consumer anxiety about AI purchases? Can a merchant-paid agent be loyal to both the merchant and the user? Does capability advantage in negotiations persist across different product categories and deal types? Does preference articulation cost explain caution about full purchase delegation? Does stripping identity cues from shopping dialogue reduce seller inference accuracy? Can a buyer agent conceal willingness to pay while still shopping naturally in character? How much does removing financial vocabulary from a consumer profile weaken willingness to pay signals?See all 44 inquiring lines on this note →
Do shopping agents respond more strongly to price and rating cues?
Explores whether AI agents manipulated by prices, ratings, and nudges shift their choices more dramatically than humans facing identical options. Matters because it affects how exploitable agents are to persuasion design.
Irreversibility and visibility, not stakes, trigger trust withdrawal Language model groups match human outcome patterns but through different processes. LLM choice biases originate in training data, not cognitive limits Costly information plus vague goals drive AI shopping heuristics, not model traits AI brand recommendations drove measurable search and visit lifts among non-customers
Why has marketing research neglected prompting behavior and relational dynamics with AI? Can a merchant-funded AI agent remain loyal to the consumer's interests? When do consumers overtrust AI agents built to represent their own preferences? Do AI shopping agents show consistent position bias in text-only interfaces? How do different AI models weigh price signals versus product description text? What conflicts exist between vendor incentives and objective consumer research findings? Will merchants deliberately exploit stable position biases to influence agent purchasing? How does delegation trust respond when AI agents make exploitable consumer choices invisibly? Can multimodal cues or larger choice sets reduce agents' rule-like decision patterns? What new vulnerabilities emerge when AI agents replace humans as purchasing decision makers? Do AI shopping agents favor items from preferred sources over objectively better alternatives? What safeguards beyond sign-off do consumers want from autonomous shopping agents?See all 91 inquiring lines on this note →
Will brands succeed by building their own AI shopping agents?
Explores whether brands can compete by creating proprietary AI agents that mediate consumer purchases, or whether structural conflicts of interest and missing ecosystem conditions will undermine this strategy.
Can an AI agent serve both merchant and user interests fairly? Why do capable AI agents still fail in real deployments? Do consumers actually want to delegate shopping decisions to AI? Do shopping agents respond more strongly to price and rating cues? Why did Amazon block Meta's shopping agent while opening its own?
Merchant-funded agents cannot loyally serve both the merchant and user. Agent capability alone cannot ensure successful real-world deployment Consumers report willingness to delegate shopping but retain control over brands Shopping agents shift choices far more than humans under identical manipulations Amazon's agent blocking is a fight over transaction fees, not AI stance
How does machine agency shift from passive tool to active partner in decisions? Why has marketing research neglected prompting behavior and relational dynamics with AI? Can a merchant-funded AI agent remain loyal to the consumer's interests? Do AI agents currently verify brand claims against reality at scale? Why do weekly AI users prefer brand-built agents over AI platforms? Why do safeguards matter more than AI capability for consumer adoption? What conflicts exist between vendor incentives and objective consumer research findings? Will merchants deliberately exploit stable position biases to influence agent purchasing? What legal standardization and trust infrastructure do AI purchasing agents require? Do AI shopping agents favor items from preferred sources over objectively better alternatives? Do consumers who trust AI agents understand their actual limitations? What safeguards beyond sign-off do consumers want from autonomous shopping agents?See all 55 inquiring lines on this note →
Do consumers actually want to delegate shopping decisions to AI?
Accenture's survey reports 74% of consumers trust AI agents with purchases, but the data reveals selective delegation bounded by brand loyalty and safeguard requirements. The question asks whether stated willingness reflects genuine appetite or reflects survey framing and vendor bias.
Does the personal assistant model actually serve most users? Why do capable AI agents still fail in real deployments? Will brands succeed by building their own AI shopping agents? Do consumers want to approve AI shopping purchases before they happen? Does naming a payments brand increase trust in AI transactions?
Personal assistant automation appeals to narrow professionals not general users Agent capability alone cannot ensure successful real-world deployment Consumers expect AI agents to reshape purchasing but infrastructure gaps remain 35 percent of consumers demand approval before AI agents complete purchases Payments brand attachment narrows the AI payment trust gap significantly
Can a merchant-funded AI agent remain loyal to the consumer's interests? When do consumers overtrust AI agents built to represent their own preferences? How does shopping task delegation differ from routine administrative delegation? Does stated trust in surveys predict actual delegated purchasing behavior? How does delegation trust respond when AI agents make exploitable consumer choices invisibly? Do consumers who trust AI agents understand their actual limitations? What percentage of shoppers constitute the super-truster segment exactly? What safeguards beyond sign-off do consumers want from autonomous shopping agents? Can a merchant-paid agent be loyal to both the merchant and the user? Does stated trust in brands predict actual delegation of agentic commerce? How does trust transfer between AI systems and established payment brands? Why do consumers resist letting AI control payment authority directly?See all 44 inquiring lines on this note →
Can AI seller agents shift buyer preferences through description edits?
This research tests whether seller-side AI agents can strategically manipulate product descriptions to influence AI shopping agents' purchasing decisions. The question matters because it reveals whether autonomous agents are vulnerable to the same optimization pressures that shaped human-facing digital markets.
Will agents compete for attention just like users do? Do AI shopping agents show persistent position bias in text interfaces? Why do AI-delegated firms stop exploring new business models? Can cheap agent search actually harm shopping markets? Does caching fares give infrastructure providers pricing power?
Agent attention economy will replace user attention competition AI shopping agents show persistent position bias resistant to prompting and interface changes AI-delegated strategy converges to profitable lock-in that discourages novelty Cheap agent search can leave markets worse off despite lower costs Amadeus caches precomputed fares to handle explosive agent-driven search volume
Can a merchant-funded AI agent remain loyal to the consumer's interests? Do AI agents currently verify brand claims against reality at scale? What happens when multiple seller agents optimize descriptions simultaneously? Can real marketplaces sustain the rapid description changes AI optimization requires? How do different AI models weigh price signals versus product description text? What conflicts exist between vendor incentives and objective consumer research findings? Will merchants deliberately exploit stable position biases to influence agent purchasing? What new vulnerabilities emerge when AI agents replace humans as purchasing decision makers? What legal standardization and trust infrastructure do AI purchasing agents require? Which LLM models consistently win better deals as both buyer and seller? Do AI shopping agents favor items from preferred sources over objectively better alternatives? What safeguards beyond sign-off do consumers want from autonomous shopping agents?See all 37 inquiring lines on this note →
Do AI shopping agents show persistent position bias in text interfaces?
This explores whether position bias—a model-specific tendency to favor products at certain locations—survives in stripped-down, text-only settings and resists corrective prompting, and what that reveals about where the bias originates.
Does a single benchmark score actually predict agent readiness? Will agents compete for attention just like users do? Can AI seller agents shift buyer preferences through description edits? When do pricing cues actually mislead AI shopping agents? Do LLM agents achieve fair marketplace outcomes when given ideal search?
Agent capability requires a vector across at least five separable axes Agent attention economy will replace user attention competition AI seller agents shifted market share by editing product descriptions alone Costly information plus vague goals drive AI shopping heuristics, not model traits LLM agents approach optimal welfare only under ideal search conditions
Why has marketing research neglected prompting behavior and relational dynamics with AI? Do AI shopping agents show consistent position bias in text-only interfaces? How do different AI models weigh price signals versus product description text? Can position bias in AI agents be mitigated through explicit instruction or prompt design? Does position bias vary predictably across different AI model families and versions? Will merchants deliberately exploit stable position biases to influence agent purchasing? How quickly does position bias change when an AI model receives a major update? Can multimodal cues or larger choice sets reduce agents' rule-like decision patterns? Do AI shopping agents favor items from preferred sources over objectively better alternatives? Do the same brands appear in different tiers across multiple AI search engines? Does visible transaction history reduce or increase consumer anxiety about AI purchases? How stable or volatile are individual brand mentions across repeated AI queries?See all 45 inquiring lines on this note →
How much traffic do AI assistants actually send to retailers?
Adobe's holiday 2025 data tracks how much retail traffic comes from generative AI referrals and whether that traffic behaves differently from other sources, revealing potential shifts in how consumers discover and purchase products.
How much are AI Overviews actually reducing organic search clicks? Will AI Overviews reduce search referral traffic to publishers? How much do shoppers trust AI shopping assistants? Does naming a payments brand increase trust in AI transactions? Does Adobe's AI referral traffic growth statistic hold up under scrutiny?
AI Overviews correlate with 58 percent lower clickthrough for top results Publishers expect search referral traffic to decline 40 to 43 percent Shoppers trust AI shopping assistants more than media and human advisors Payments brand attachment narrows the AI payment trust gap significantly Adobe's Gen AI traffic growth claim masked a negligible underlying base
Why are holiday returns declining as AI-referred traffic increases? What affiliate fee rates do different merchants currently offer for referrals? How do headline growth percentages obscure small absolute changes in traffic? Can Gen AI referral traffic grow while search engine referral traffic simultaneously declines?
Does model capability determine who wins in agent negotiations?
When LLM agents negotiate prices with each other, does raw model capability consistently predict better deal outcomes, or do negotiation strategies matter more? This matters because it could reveal a new form of economic disparity if agentic commerce scales.
Does a single benchmark score actually predict agent readiness? Why do AI-delegated firms stop exploring new business models? Where do user values break down in agent supervision? Do LLM negotiation agents reliably follow budget and pricing rules? Can AI agents designed by competing principals destroy bargaining surplus?
Agent capability requires a vector across at least five separable axes AI-delegated strategy converges to profitable lock-in that discourages novelty User values align with agent output but fracture during supervision LLM negotiation agents frequently violate budget constraints and overpay regardless of model. Flexible AI agent preferences risk a race to the bottom destroying surplus
Which LLM models consistently win better deals as both buyer and seller? Does capability advantage in negotiations persist across different product categories and deal types? How does worker bargaining power connect to long-term wage growth? Do frontier model differences disappear when agents can swap LLMs? How does marketplace scale impact LLM agent decision quality and efficiency? Do compute-funded negotiation strategies compound with model capability advantages? Does model size heterogeneity break or stabilize LLM collusion? Why do LLM agents focus on prices rather than reasoning about incentives? Do more capable LLMs exploit weaker agents consistently across market types? Does capability asymmetry between negotiating agents produce unfair market outcomes?
Why does manual search fail as product choices become complex?
As product variety grows in high-dimensional preference spaces, manual browsing hits a wall. Does shifting to agent-assisted search with refined preference signals solve that collapse?
Agent attention economy will replace user attention competition Ten adaptive questions suffice to infer personalized reward coefficients Platforms want agentic search before consumers rationally should switch Costly information plus vague goals drive AI shopping heuristics, not model traits AI agents will disrupt shopping through four staged evolutions of control transfer
Why do LLM agents collapse into deterministic rules under single-attribute product comparisons? How much do AI recommendation lists vary when you run the same search twice? How does agentic search shift scarcity from preference discovery to articulation? What makes preference articulation harder than browsing in high-dimensional spaces? Can hybrid shopping journeys combine manual inspection with agentic refinement? How should platforms invest if preference signal quality matters more than catalog breadth? Why might users over-trust preference agents despite rational adoption hesitation? Why does preference articulation replace manual comparison in agentic search? What makes preference elicitation irreducibly vague in some cases? Does preference articulation become scarcer as search costs fall? Why would a platform prefer to control basket assembly rather than just product search? Why do AI shopping agents rely on marketing heuristics when looking up product attributes?See all 17 inquiring lines on this note →
Why do consumers delay adopting agentic search?
Platforms benefit from agentic search before consumers do privately. This note explores whether rational cost-minimization—weighing time spent on search itself—explains adoption delays beyond simple distrust of AI.
Why does manual search fail as product choices become complex? Does co-design participation hide misalignment in preference agents? Why do AI-delegated firms stop exploring new business models? Could agent-to-agent communication reshape market power? Who captures value when AI agents cut coordination costs?
Agentic search escapes complexity collapse by prioritizing preference articulation Co-design participation may mask agent misalignment through systematic overtrust AI-delegated strategy converges to profitable lock-in that discourages novelty Agent communication could redistribute market power by lowering switching costs Agent platforms concentrate value despite narrower price dispersion
Does stated trust in brands predict actual delegation of agentic commerce? Can platforms use lower AI fidelity to nudge consumers toward adoption? How does agentic search shift scarcity from preference discovery to articulation? Why might users over-trust preference agents despite rational adoption hesitation? How does search cost differ from the cost of changing what an AI agent optimizes for? Why do consumer protection frameworks designed for humans break down with algorithmic agents? What organizational adoption barriers exist beyond affordability and accessibility? Can agent platforms profit more from search than consumers benefit from it? Does competitive pressure force AI firms to rush agentic adoption? Why do AI agents generate vastly more searches than human travelers? Why does preference articulation replace manual comparison in agentic search? What consumer and merchant benefits does agentic commerce create?See all 36 inquiring lines on this note →
Do AI assistant brand mentions drive real user behavior?
When conversational AI names a brand to someone unfamiliar with it, does that mention actually change what the person searches for or visits online? The question matters because current analytics miss these upstream touchpoints entirely.
How consistent are AI brand recommendation lists across repeated prompts? Do AI search engines favor well-known brands over others? Do shopping agents respond more strongly to price and rating cues? Does limiting product choices increase trust in voice assistants? Does pooling AI brand mentions mask the true recommendation effect?
AI brand lists repeat less than one percent of the time in identical order AI search engines cite brands almost entirely by existing stature Shopping agents shift choices far more than humans under identical manipulations Fewer product alternatives evaluated increased satisfaction and willingness to delegate to voice assistants Pooling all AI brand mentions overstates the recommendation effect by two-fold.
How do citation rates vary across brands in different AI search engines? How often do AI assistants recommend the same brand across different conversations? Why do standard web analytics fail to credit brand mentions in chat? Do AI search engines cite brands equally regardless of market tier or status? Can transparency about AI sources backfire if brands carry party associations?
What identity standards do shopping agents actually need?
Explores whether AI shopping agents require standard protocols for identifying themselves to retailers and proving user authorization, or if existing security measures suffice.
Can a protocol split agent access from agent incentives? Can an AI agent serve both merchant and user interests fairly? Can credential vaulting alone solve merchant-facing agent identification? Why did Amazon block Meta's shopping agent while opening its own? Do broad consent settings pose bigger risks than prompt injection attacks?
Personal Agent Protocol governs access control, not agent incentive conflicts Merchant-funded agents cannot loyally serve both the merchant and user. Vaulted credentials protect consumers but do not establish merchant-facing identification Amazon's agent blocking is a fight over transaction fees, not AI stance Broad consent settings pose greater risks than prompt injection in shopping agents
Does Personal Agent Protocol solve the identity and authorization gap for shopping agents? Can OAuth-based access grants work without adding an explicit agent self-identification layer? How should agent operators demonstrate both security features and identification signals separately? What identity and authorization standard would satisfy both consumer and merchant requirements? Why does Amazon's objection focus on visibility and consent rather than credential safety? Can courts decide whether delegating your shopping to an AI agent is legally permissible? Could a cross-industry standard let users control what access agents have? Do consumers understand what access permissions they grant to personal agents? What specific concerns prevent consumers from granting AI agents complete purchase autonomy? Can retailers use off-site agents to compete without losing customer loyalty? Why do shoppers trust retailer-built agents more than third-party AI agents? Can authenticated delegation tokens prevent merchants from absorbing agent payment liability?See all 58 inquiring lines on this note →
Can credential vaulting alone solve merchant-facing agent identification?
Does securing consumer passwords through vaulted storage address retailers' need to know which automated system is making requests and whether they consented to serve it?
What identity standards do shopping agents actually need? Can a protocol split agent access from agent incentives? Can an AI agent serve both merchant and user interests fairly? Do broad consent settings pose bigger risks than prompt injection attacks? What blocks autonomous agents from handling purchases at scale?
Shopping agents lack merchant-facing identity and authorization protocols Personal Agent Protocol governs access control, not agent incentive conflicts Merchant-funded agents cannot loyally serve both the merchant and user. Broad consent settings pose greater risks than prompt injection in shopping agents Authentication, liability, and fraud are the core barriers to scaling agent-delegated commerce
Does Personal Agent Protocol solve the identity and authorization gap for shopping agents? What identity and authorization standard would satisfy both consumer and merchant requirements? Why does Amazon's objection focus on visibility and consent rather than credential safety? Could a cross-industry standard let users control what access agents have? Do consumers understand what access permissions they grant to personal agents? Can authenticated delegation tokens prevent merchants from absorbing agent payment liability? How might card networks recognize agent-operator liability to shift risk away from merchants? What fraud losses result when behavioral biometrics disappear in agent transactions? Are merchants actually adopting these protocols as an integrated stack? How does merchant-of-record status persist through API-based agent integrations? What counts as sensitive information beyond credit card sharing? Why do independent hotels depend on vendor support rather than direct MCP deployment?See all 19 inquiring lines on this note →
Who captures value when AI agents cut coordination costs?
As AI agents automate search, negotiation, and comparison work, will consumers gain lower prices and wider competition, or will agent platforms themselves capture the savings by sitting between buyers and sellers?
Coordination architectures trade production, coordination, and vulnerability costs differently Agent attention economy will replace user attention competition Platforms want agentic search before consumers rationally should switch AI agents will dramatically reduce transaction costs and redraw firm boundaries Platform pricing of inspection tokens induces excess search and seller harm
Does caching fares give infrastructure providers pricing power?
As AI agents drive search volume toward 200,000 queries per ticket, travel platforms like Amadeus are caching fares rather than pricing each query live. Does this caching layer give whoever controls it real influence over which fares agents recommend and how airlines compete?
Can AI seller agents shift buyer preferences through description edits? Why do consumers delay adopting agentic search? Are AI agents actually disintermediating online travel platforms? Are AI agents making attention-based ads irrelevant for aggregators? Are AI agents actually removing online travel agencies from bookings?
AI seller agents shifted market share by editing product descriptions alone Platforms want agentic search before consumers rationally should switch OTA stock gains suggest agents commoditize discovery without disintermediating fulfillment AI agents strip aggregators of ad relevance by keeping the demand signal AI agents are embedding OTAs deeper as settlement infrastructure, not removing them
How does caching refresh cadence affect fare accuracy for consumers? Can platforms tune caching refresh rates to their own economic benefit? Does infrastructure control over fares shift market power away from airlines? Why is Booking Holdings' marketing expense ratio rising faster than gross bookings? How do booking aggregators like Booking.com justify their 14-15 percent take rate? Can centralized data ownership alone move travel agents out of pilot mode? Does MCP integration guarantee accurate rate and availability data at booking scale? Where exactly does fulfillment infrastructure value concentrate in travel booking?
Where does an AI assistant's real value actually come from?
Does an AI assistant help by providing better information, or by actually executing actions the consumer already approved but failed to do themselves? This matters for understanding what value persists when markets adapt.
Sellers recapture one third of AI assistant savings through headline repricing Platforms want agentic search before consumers rationally should switch EU consumer law's human-centric design fails for algorithmic consumer agents AI agents will dismantle negative option billing by eliminating consumer forgetfulness Cheap agent search can leave markets worse off despite lower costs
Does platform pricing of inspection tokens encourage excess search?
Explores whether profit-maximizing platforms that price access to buyer attention systematically over-engineer inspection relative to what benefits society. This matters because it reveals how platform incentives can distort market efficiency even when consumer welfare stays flat.
Seller subsidies sort quality with top firms pooling at the cap Platforms want agentic search before consumers rationally should switch Agentic search escapes complexity collapse by prioritizing preference articulation Algorithmic recommendation advantage shrinks as seller competition intensifies Agent platforms concentrate value despite narrower price dispersion
Can agent platforms profit more from search than consumers benefit from it? Can platforms tune caching refresh rates to their own economic benefit? Does infrastructure control over fares shift market power away from airlines? Does preference articulation become scarcer as search costs fall? Why would a platform prefer to control basket assembly rather than just product search? Why might agentic search become profitable for platforms before helping consumers? Why is Booking Holdings' marketing expense ratio rising faster than gross bookings? How are platforms like DoorDash building hundred-billion-dollar ad businesses on the human attention layer? How do booking aggregators like Booking.com justify their 14-15 percent take rate? Can platforms profit from agentic search before consumers actually prefer it? What makes protocols more market-shaping than APIs enforced by dominant actors? Can de facto protocols prevent market re-concentration after initial adoption?See all 31 inquiring lines on this note →
Can an algorithm recommend products without matching an ideal mechanism?
Do simple recommendation rules that avoid transfers and forced purchases actually achieve the same welfare outcomes as optimal direct mechanisms that can demand payment and specify allocations?
Why does manual search fail as product choices become complex? Why do consumers delay adopting agentic search? Does algorithmic recommendation lose its edge as sellers compete more? Does platform pricing of inspection tokens encourage excess search? How do seller subsidies sort quality in search markets?
Agentic search escapes complexity collapse by prioritizing preference articulation Platforms want agentic search before consumers rationally should switch Algorithmic recommendation advantage shrinks as seller competition intensifies Platform pricing of inspection tokens induces excess search and seller harm Seller subsidies sort quality with top firms pooling at the cap
Why would a platform prefer to control basket assembly rather than just product search? What would happen if users discovered their agent's recommendations were paid for? Would separating agent recommendations from merchant payments reduce principal-agent risk? How does subsidy signaling compare to auction-based ranking mechanisms? Why does platform revenue motive conflict with consumer welfare in search markets? Does algorithmic recommendation welfare depend on whose surplus the agent protects? How does this mechanism differ from commitment power in repeated buyer-seller interactions? Does the beneficiary fix work outside hotel booking scenarios? Can an optimal algorithm match what a direct mechanism achieves? How do sellers strategically respond to algorithmic recommendation design? Can assistance tools be designed to encourage adoption of welfare-maximizing modalities?
Do AI agents actually benefit from network effects?
Chen explores whether consumer AI agents can achieve winner-take-all network effects or whether interoperability and swappable underlying models prevent lock-in. This matters because it determines whether agent markets will consolidate or remain fragmented.
Competitive pressure creates collective-action failures in agentic AI safety vetting Receiving misaligned email raises odds of sending one by forty-two percent Agent communication could redistribute market power by lowering switching costs Coordination and common causes require distinct causal evidence to separate AI agents send misaligned messages in most long-horizon market runs
What consumer and merchant benefits does agentic commerce create? Does mixing agent types reduce unexplored product alternatives in markets? Can AI layers on TMC platforms change underlying OTA supply relationships? Why do incumbent platforms retain advantages under agentic commerce systems? Can open agent webs still concentrate economic benefits despite avoiding walled gardens? Why do tech incumbents have structural advantages in controlling agent interoperability? What five conditions must exist for agent ecosystems to generate value? Which markets have intermediaries that agents could most easily replace first? Can brands compete to become an AI agent's preferred choice? Can brands compete effectively if agents become the primary consumer decision layer?
Do smarter AI agents negotiate better deals than weaker ones?
Explores whether a model's general capability translates to real negotiation advantage when representing human traders. Matters because it could shape how AI agents handle high-stakes commerce.
Does model capability determine who wins in agent negotiations? Do people notice when their agents secure worse deals? Do agent representation failures or negotiation skills cause market shortfalls? Does model capability or instructions shape agent negotiation outcomes? Do LLM agents achieve fair marketplace outcomes when given ideal search?
More capable LLMs consistently secure better negotiation deals as buyers and sellers People did not notice they received worse deals from weaker agents Representation failures, not negotiation skill, caused most market shortfall Model capability outweighs instructions in shaping market efficiency LLM agents approach optimal welfare only under ideal search conditions
How do human traders adjust their behavior when facing different AI capabilities? Do autonomous agents start with low-risk purchases before handling high-stakes decisions? Does the negotiation gap between models persist outside controlled simulations? How do satisfaction ratings differ from objective outcomes in delegated negotiations? Can self-report feedback detect agent quality gaps in real markets? Do stronger language models win better deals in pairwise negotiations? How accurately do AI agents represent human preferences in markets? Do monitoring checkpoints before trading improve agent representation of preferences? What makes agent-to-agent negotiation fundamentally different from agent-assisted shopping? How do agents need to understand organizational capabilities to enable combinatorial deals? What prevents human-run markets from discovering valuable multi-party transaction combinations? Does first-proposal bias outweigh capability differences in agent negotiation outcomes?See all 30 inquiring lines on this note →
Do people notice when their agents secure worse deals?
When weaker AI agents negotiate on people's behalf, do the people realize they got inferior prices? This matters for detecting unfair outcomes in future agent-driven markets where people may lack visibility into actual results.
Do smarter AI agents negotiate better deals than weaker ones? Does model capability determine who wins in agent negotiations? Where do user values break down in agent supervision? Do agent representation failures or negotiation skills cause market shortfalls? Do LLM agents achieve fair marketplace outcomes when given ideal search?
Stronger Claude models secure measurably better deals as both buyers and sellers More capable LLMs consistently secure better negotiation deals as buyers and sellers User values align with agent output but fracture during supervision Representation failures, not negotiation skill, caused most market shortfall LLM agents approach optimal welfare only under ideal search conditions
How do human traders adjust their behavior when facing different AI capabilities? How do satisfaction ratings differ from objective outcomes in delegated negotiations? Can self-report feedback detect agent quality gaps in real markets? What supervision methods help humans notice unfair agent-mediated deals? Why would merchants accept higher agent take rates as they mature? Why do most people distrust AI agents with payments? How might principals instruct agents to balance speed against better deal terms? Can price discrimination bans work without restructuring how agents are financed? Can sellers credibly signal quality through inspection cost subsidies alone? What does epistemic fairness require for AI agents evaluating paid placements? Does capability asymmetry between negotiating agents produce unfair market outcomes?
Do agent representation failures or negotiation skills cause market shortfalls?
When AI agents trade on behalf of people in markets, does the gap from ideal outcomes stem from agents misunderstanding what their person wants, or from poor negotiation once trading begins? Understanding this matters for where to focus improvements.
Do shopping agents respond more strongly to price and rating cues? Can a seller infer what a buyer really wants to pay? Must AI monitoring precede incentive alignment as agents evolve? Does model capability or instructions shape agent negotiation outcomes? Do agent mistakes or misaligned incentives cause marketplace welfare loss?
Shopping agents shift choices far more than humans under identical manipulations Sellers infer willingness to pay nearly perfectly from shopping dialogue alone AI agency requires scaffolded monitoring before incentive alignment works Model capability outweighs instructions in shaping market efficiency Agent welfare loss stems from execution mistakes not misaligned incentives
How do human traders adjust their behavior when facing different AI capabilities? What systematic failures occur when LLM agents negotiate market constraints? How might principals instruct agents to balance speed against better deal terms? Why do AI agents infer beneficiary from their employer? Does capability asymmetry between negotiating agents produce unfair market outcomes? Can a single AI agent act as a market maker in multi-party negotiation?
Does model capability or instructions shape agent negotiation outcomes?
When AI agents trade in multi-party markets, does swapping the underlying model or changing their behavioral instructions have a bigger effect on trading results? This matters for designing markets that work efficiently and fairly.
Does model capability determine who wins in agent negotiations? Do LLM negotiation agents reliably follow budget and pricing rules? Do agent representation failures or negotiation skills cause market shortfalls? Do smarter AI agents negotiate better deals than weaker ones? Do agent mistakes or misaligned incentives cause marketplace welfare loss?
More capable LLMs consistently secure better negotiation deals as buyers and sellers LLM negotiation agents frequently violate budget constraints and overpay regardless of model. Representation failures, not negotiation skill, caused most market shortfall Stronger Claude models secure measurably better deals as both buyers and sellers Agent welfare loss stems from execution mistakes not misaligned incentives
How do human traders adjust their behavior when facing different AI capabilities? Do frontier model differences disappear when agents can swap LLMs? Can prosocial instructions reduce AI agent self-interest in trading? How do ownership structures shape whether agents favor sellers or buyers? What signals beyond price and ratings actually move AI agent choices? How might principals instruct agents to balance speed against better deal terms? What happens when one party controls vastly more negotiation compute? How does pre-commitment power reshape what algorithms can achieve in two-sided markets? Does capability asymmetry between negotiating agents produce unfair market outcomes? Does delegating negotiation to AI with tunable preferences destroy bargaining surplus? Can a single AI agent act as a market maker in multi-party negotiation?
Can better tools fix an AI agent's exploitable judgment?
Project Vend tested whether scaffolding and oversight could make Claude-based shopkeeper Claudius both more effective and more reliable at avoiding naive mistakes like illegal contracts or security misjudgments.
AI collaboration boosts output but risks eroding oversight skills Anthropic reports AI systems handling more development work with rising speed. Anthropic ranks trust holistically rather than as a strict operational hierarchy Heavy Claude delegators report greater career optimism than light users Domain expertise, not coding skill, predicts agentic-coding success
Why do complex tasks show the least oversight when Claude struggles most? What disclosure or auditing could make merchant-funded agents trustworthy? Can procedural guardrails prevent AI agents from making naive mistakes? Can operators reduce Claude's trust through their system prompts? How does Claude detect when operator instructions become clearly harmful?
Do AI models follow budgets more rigidly than humans do?
Explores whether AI systems treat numerical constraints as absolute rules while humans treat them as flexible guidelines, and what mechanisms might explain the gap.
Does machine agency exist on a spectrum rather than binary? Does AI personality fit matter more than AI presence itself? Do shopping agents respond more strongly to price and rating cues? Do AI shopping agents reshape what consumers actually want? Do human shopping tactics work the same way on AI agents?
Machine agency operates on a five-level spectrum from passive to cooperative AI personality matched to human traits shifted creative collaboration outcomes Shopping agents shift choices far more than humans under identical manipulations Transactional AI agents co-construct consumer preferences through repeated conversation loops Star ratings and price consistently sway AI agents, but scarcity and bundles do not
How does the cost of information acquisition influence AI agent decision-making behavior? Can AI agents consistently remain objective under different revenue models? How can systems separate AI decisions from actual payment execution? Can product embedding overcome user discomfort with AI controlling finances? Can revenue pressure change how AI systems rank or phrase responses? How do autonomy budgets map onto OpenAI's Model Spec scope requirements? What interaction mechanisms help systems determine when to defer to humans? What monthly spending caps are banks actually imposing on frontier AI models? How do token-count variations affect cost estimates as AI workflows become more agentic? Does AI cost reduction through model routing translate to actual labor displacement? What cost governance strategies have banks successfully deployed for AI? Why do humans override or ignore AI proposals even when they increase joint surplus?
Does AI personality fit matter more than AI presence itself?
When humans collaborate with AI on creative tasks, does matching the AI's personality traits to the human's own characteristics actually improve results, or does the productivity gain come simply from working with AI at all?
Configuration files make live human-AI protocols auditable and shareable across labs Human-AI teams underperform the best alone except in creation tasks Generative AI matched individual performance to teams without AI at Procter & Gamble AI models reject small budget overages humans accept until shown the reasoning AI agents are passive by design, not by capability
Do broad consent settings pose bigger risks than prompt injection attacks?
Does the real danger in shopping agents come not from adversarial attacks but from overly permissive one-time authorization checkboxes that override per-transaction safeguards?
Do consumers want to approve AI shopping purchases before they happen? Can a protocol split agent access from agent incentives? Why did Amazon block Meta's shopping agent while opening its own? Can credential vaulting alone solve merchant-facing agent identification? What identity standards do shopping agents actually need?
35 percent of consumers demand approval before AI agents complete purchases Personal Agent Protocol governs access control, not agent incentive conflicts Amazon's agent blocking is a fight over transaction fees, not AI stance Vaulted credentials protect consumers but do not establish merchant-facing identification Shopping agents lack merchant-facing identity and authorization protocols
Should retailers be able to refuse agents before any commercial relationship begins? Does Personal Agent Protocol solve the identity and authorization gap for shopping agents? Why does Amazon's objection focus on visibility and consent rather than credential safety? Does an agent paid by merchant referral fees face a loyalty conflict? Do consumers understand what access permissions they grant to personal agents? What explains gaps between advertised safeguards and setup-time permission mechanisms? Why do agent launch announcements differ from what one-time consent actually enables? Do permissioning frameworks resolve merchant accountability for agent ordering errors? Why do Chinese wallets concentrate specifically in security services on chain? Why do consumers want approval rights before agents complete purchases? Why do consumers want to approve purchases before agents pay? What counts as sensitive information beyond credit card sharing?See all 24 inquiring lines on this note →
Why did Amazon block Meta's shopping agent while opening its own?
Amazon blocked Meta's Muse from shopping on its platform while running its own cross-site buying agent. Is this really about AI adoption, or about who controls the transaction and collects fees?
Can an AI agent serve both merchant and user interests fairly? Can a protocol split agent access from agent incentives? Will brands succeed by building their own AI shopping agents? Why did Amazon and Shopify respond so differently to AI agents? What identity standards do shopping agents actually need?
Merchant-funded agents cannot loyally serve both the merchant and user. Personal Agent Protocol governs access control, not agent incentive conflicts Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Advertising dependence explains Amazon's agent blockade versus Shopify's openness Shopping agents lack merchant-facing identity and authorization protocols
Can an AI agent serve both merchant and user interests fairly?
Explores whether an agent funded by merchant referral fees can make unbiased recommendations on behalf of users, or whether the payment structure creates an unavoidable conflict of interest.
Agent attention economy will replace user attention competition Personal assistant automation appeals to narrow professionals not general users Advertising dependence explains Amazon's agent blockade versus Shopify's openness Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Shopping agents lack merchant-facing identity and authorization protocols
How will merchant fees shape competition for agent invocations rather than user clicks? Why does Amazon block Muse but run its own agents elsewhere? What disclosure or auditing could make merchant-funded agents trustworthy? What happens to publisher revenue when search referral traffic collapses? Do smaller ad-dependent publishers face larger traffic losses from AI summaries? How do commercial incentives shape vendor claims about AI and collaboration? How do courts assign liability when AI intermediaries cause harm to consumers? How much referral traffic do ChatGPT and Google each send to publishers? Can AI companies mobilize users as advocates like Uber or Airbnb did? Can AI systems distinguish between users' stated preferences and their genuine long-term interests? Can a merchant-funded AI agent remain loyal to the consumer's interests? When do consumers overtrust AI agents built to represent their own preferences?See all 122 inquiring lines on this note →
Could AI agents reassemble shopping baskets across multiple retailers?
Bain & Company explores whether AI agents that compare prices and products across vendors could fragment the traditional retailer-controlled basket, especially in grocery. This matters because it would shift competitive pressure from individual products to entire baskets.
Platforms want agentic search before consumers rationally should switch Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Three AI agent types reshape retail with different competitive stakes Agent platforms concentrate value despite narrower price dispersion Agent ownership determines whose interests the agent optimizes for
What happens to retailer margins if price comparison extends to entire shopping baskets? How would becoming an agent of choice protect brands in a multivendor basket world? Can retailers use off-site agents to compete without losing customer loyalty? Why would a platform prefer to control basket assembly rather than just product search? Which retail categories are most vulnerable to agent-driven basket fragmentation?
How are different AI agents reshaping the retail shopper journey?
As AI shopping agents proliferate, are they organized into distinct types with different competitive impacts? Understanding these categories matters because each poses different threats and opportunities for traditional retailers.
Will brands succeed by building their own AI shopping agents? Do consumers want to approve AI shopping purchases before they happen? Could AI agents reassemble shopping baskets across multiple retailers? Do advertising-funded shopping agents favor paid product placements? Does agent ownership determine whose interests get optimized?
Consumers expect AI agents to reshape purchasing but infrastructure gaps remain 35 percent of consumers demand approval before AI agents complete purchases Multivendor agent baskets could dissolve retailer control and intensify price competition Advertising-funded shopping agents risk objective bias from placement incentives Agent ownership determines whose interests the agent optimizes for
How will AI agents progressively take over online shopping?
BCG proposes that AI will disrupt e-commerce through four staged evolutions, gradually transferring control from humans to agents across discovery, checkout, product selection, and payment optimization. Understanding this staging helps anticipate where disruption hits first.
Bilgihan's framework splits AI agents into advisory and transactional types Agentic search escapes complexity collapse by prioritizing preference articulation Consumers and networks gain most from agentic commerce Platforms want agentic search before consumers rationally should switch Payment systems need three layers separating agent decisions from settlement
No Inquiring Lines yet
Who wins and loses as AI agents reshape shopping?
A BCG survey explores how agentic commerce redistributes value across the shopping ecosystem. The research measures consumer adoption intent, spending impact, and which parties benefit most from autonomous purchasing.
Do consumers actually want to delegate shopping decisions to AI? Do consumers want to approve AI shopping purchases before they happen? What drives shoppers to accept autonomous AI assistants? How will AI agents progressively take over online shopping? Does naming a payments brand increase trust in AI transactions?
Consumers report willingness to delegate shopping but retain control over brands 35 percent of consumers demand approval before AI agents complete purchases Perceived benefits drive autonomous AI acceptance under scarcity conditions AI agents will disrupt shopping through four staged evolutions of control transfer Payments brand attachment narrows the AI payment trust gap significantly
What consumer and merchant benefits does agentic commerce create? What product categories are consumers most ready to let AI agents buy autonomously? How would becoming an agent of choice protect brands in a multivendor basket world? What happens to traditional retailer-shopper relationships as third-party agents grow? Which retail categories are most vulnerable to agent-driven basket fragmentation? Does merchant liability move up the agentic commerce protocol stack? How does agentic commerce on crypto differ from consumer-authorized AI purchases? Why do incumbent platforms retain advantages under agentic commerce systems? How do merchants win selection when agents control the purchase decision? Can brands survive if they are not the agent of choice to consumers? Why does adoption urgency for agentic commerce exceed its near-term financial opportunity? How do margin assumptions change the retail media versus agentic commerce comparison?See all 20 inquiring lines on this note →
Do AI agents actually reduce friction in electronic markets?
As autonomous agents handle search and negotiation, do they eliminate traditional market frictions through cheaper information processing, or do they create new behavioral and informational obstacles instead?
Do shopping agents respond more strongly to price and rating cues? Do AI shopping agents show persistent position bias in text interfaces? Why does manual search fail as product choices become complex? Does agent diversity actually improve market competition? Can cheap agent search actually harm shopping markets?
Shopping agents shift choices far more than humans under identical manipulations AI shopping agents show persistent position bias resistant to prompting and interface changes Agentic search escapes complexity collapse by prioritizing preference articulation Agent diversity partially offsets bias but incompletely improves competition Cheap agent search can leave markets worse off despite lower costs
Does agent diversity actually improve market competition?
When AI agents with different biases trade in the same market, do their differences cancel out individual search limitations? Understanding this matters for whether agentic markets self-correct or compound inefficiencies.
Do AI shopping agents show persistent position bias in text interfaces? Does model capability determine who wins in agent negotiations? Does higher purchase volume make AI shoppers search less carefully? Do AI agents actually reduce friction in electronic markets? Do LLM agents achieve fair marketplace outcomes when given ideal search?
AI shopping agents show persistent position bias resistant to prompting and interface changes More capable LLMs consistently secure better negotiation deals as buyers and sellers Higher purchase volume made Gemini agents abandon unit-price for simpler price minimization AI agents transform market frictions rather than eliminate them LLM agents approach optimal welfare only under ideal search conditions
How do human traders adjust their behavior when facing different AI capabilities? Why might inequality remain invisible in fully agentic economies? What happens to market entry and product variety when generation costs drop? How much agent architectural diversity exists in actual deployed marketplaces? Does mixing agent types reduce unexplored product alternatives in markets? Does bilateral agent commerce expand the discoverable frame more than unilateral agent search? Does price-war-concern mechanism work in oligopolies with more than two agents? Why do incumbent platforms retain advantages under agentic commerce systems? Can price discrimination bans work without restructuring how agents are financed? Can smaller merchants compete if agentic commerce becomes platform-specific? What role does easy market entry play in adverse selection? Why might agentic markets underperform compared to traditional consumer search markets?See all 20 inquiring lines on this note →
Do AI shopping agents reshape what consumers actually want?
Explores whether transactional AI agents that execute purchases on consumers' behalf co-construct preferences through ongoing conversation, rather than consumers forming independent preferences beforehand.
Is AI shifting from message conduit to active conversation participant? Does machine agency exist on a spectrum rather than binary? Does co-design participation hide misalignment in preference agents? How do advisory and transactional AI agents differ in function? Do AI models follow budgets more rigidly than humans do?
HCI research shows AI shifting from passive channel to active communication participant Machine agency operates on a five-level spectrum from passive to cooperative Co-design participation may mask agent misalignment through systematic overtrust Bilgihan's framework splits AI agents into advisory and transactional types AI models reject small budget overages humans accept until shown the reasoning
Can an AI agent's limited expressiveness flatten the preferences it claims to represent? How does machine agency shift from passive tool to active partner in decisions? Why has marketing research neglected prompting behavior and relational dynamics with AI? Can a merchant-funded AI agent remain loyal to the consumer's interests? When do consumers overtrust AI agents built to represent their own preferences? What happens when multiple seller agents optimize descriptions simultaneously? Why do weekly AI users prefer brand-built agents over AI platforms? Can real marketplaces sustain the rapid description changes AI optimization requires? What conflicts exist between vendor incentives and objective consumer research findings? Does stated trust in surveys predict actual delegated purchasing behavior? Why do LLM agents collapse into deterministic rules under single-attribute product comparisons? Can multimodal cues or larger choice sets reduce agents' rule-like decision patterns?See all 110 inquiring lines on this note →
How do advisory and transactional AI agents differ in function?
Bilgihan and colleagues propose distinguishing AI agents by whether they recommend options or execute decisions. Understanding this split matters for designing agent-based systems and predicting how consumers will interact with different agent types.
What UX principles do workplace users want in AI agents? Does machine agency exist on a spectrum rather than binary? Do AI shopping agents reshape what consumers actually want? How will AI agents progressively take over online shopping? Do productivity agents actually protect users better than emotional ones?
Workplace users prioritize control, reliability, context-awareness, and safety in AI agent design Machine agency operates on a five-level spectrum from passive to cooperative Transactional AI agents co-construct consumer preferences through repeated conversation loops AI agents will disrupt shopping through four staged evolutions of control transfer Productivity agents score highest, emotional agents lowest on interaction principles
How does machine agency shift from passive tool to active partner in decisions? How does the five-level agency spectrum relate to advisory versus transactional types? How do advisory agents differ from transactional agents in shopping? How can systems separate AI decisions from actual payment execution? How does agent intent differ from consumer account recognition?
Does agentic commerce shift fraud liability onto merchants?
When AI agents handle purchases, behavioral fraud signals disappear and merchants remain liable. Does this structural shift leave merchants without traditional fraud-detection protections they once relied on?
How can AI agents prove who authorized their actions? Does naming a payments brand increase trust in AI transactions? Can a protocol split agent access from agent incentives? Why do agents fail at identity verification and authorization? Does agent-initiated commerce break existing payment infrastructure?
Delegation tokens extend OAuth 2.0 to anchor AI agent actions to human principals Payments brand attachment narrows the AI payment trust gap significantly Personal Agent Protocol governs access control, not agent incentive conflicts Agent coordination requires cryptographic identity and protocol-level authorization enforcement Agent commerce exposes structural gaps in payments infrastructure designed for humans
What consumer and merchant benefits does agentic commerce create? Can authenticated delegation tokens prevent merchants from absorbing agent payment liability? How might card networks recognize agent-operator liability to shift risk away from merchants? What fraud losses result when behavioral biometrics disappear in agent transactions? Does merchant liability move up the agentic commerce protocol stack? Can existing unfair commercial practices law cover agentic transactions today? How does agentic commerce on crypto differ from consumer-authorized AI purchases? Can existing consumer protection laws address algorithmic shopping agents? Why would merchants accept higher agent take rates as they mature? Does existing consumer protection law already provide tools for AI agent oversight? Why do most people distrust AI agents with payments? How do payment and trust functions protect platform value during agent shifts?See all 22 inquiring lines on this note →
Do agentic commerce protocols compete or complement each other?
Explores whether emerging payment protocols like ACP, AP2, and UCP represent a competitive protocol war or form complementary layers addressing different risk management needs in agent-driven commerce.
Agentic commerce shifts fraud liability to merchants as fraud signals vanish NIST proposes reusing existing identity standards for agentic systems Delegation tokens extend OAuth 2.0 to anchor AI agent actions to human principals Visa's protocol uses cryptographic signatures to verify trusted AI agents McKinsey projects one trillion dollars in US agentic commerce by 2030
Does Personal Agent Protocol solve the identity and authorization gap for shopping agents? What identity and authorization standard would satisfy both consumer and merchant requirements? What consumer and merchant benefits does agentic commerce create? How do protocol-level access controls capture transaction-by-transaction authorization preferences? Can authenticated delegation tokens prevent merchants from absorbing agent payment liability? How might card networks recognize agent-operator liability to shift risk away from merchants? Why are new protocols being built instead of extending OAuth and OIDC? Does merchant liability move up the agentic commerce protocol stack? How do web browsing agents differ from protocol-native agents legally? What protocol-layer obligations could replace GUI consumer protection rules? Are merchants actually adopting these protocols as an integrated stack? Can existing unfair commercial practices law cover agentic transactions today?See all 52 inquiring lines on this note →
Can EU consumer law rules survive agent-to-agent API transactions?
Explores whether graphical interface requirements in EU consumer protection law remain enforceable when autonomous agents transact directly through APIs instead of clicking buttons. This matters because agentic commerce is shifting transactions away from human-readable interfaces.
Can EU consumer law protect AI agents as algorithmic buyers? Can traditional governance tools actually control AI agents? Will brands succeed by building their own AI shopping agents? Does agentic commerce shift fraud liability onto merchants? Should AI shopping agents face fiduciary duties like human brokers?
EU consumer law's human-centric design fails for algorithmic consumer agents Conventional governance tools fail for opaque, fast-moving AI agents Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Agentic commerce shifts fraud liability to merchants as fraud signals vanish Owens and Law argue AI agents need fiduciary duties before markets harden
What identity and authorization standard would satisfy both consumer and merchant requirements? Can courts decide whether delegating your shopping to an AI agent is legally permissible? What consumer and merchant benefits does agentic commerce create? Do consumers understand what access permissions they grant to personal agents? Does merchant liability move up the agentic commerce protocol stack? What protocol-layer obligations could replace GUI consumer protection rules? Can existing unfair commercial practices law cover agentic transactions today? When should EU regulators translate design rules into protocol standards? How do existing consumer protection laws handle algorithmic agents as decision-makers? Can existing consumer protection laws address algorithmic shopping agents? What is the regulatory gap in Google's personalized pricing rollout? How does merchant-of-record status persist through API-based agent integrations?See all 27 inquiring lines on this note →
Are AI agent wallets becoming genuinely operational instead of experimental?
Chainalysis tracks whether funding-to-spend timing and repeat-funding patterns on x402 reflect a shift from testing to habitual, purposeful agentic commerce use.
Payments brand attachment narrows the AI payment trust gap significantly Only 29% of people trust AI agents with bank account access American wallets lead AI model inference payments while China dominates security Agents will adopt cards first, then stablecoin-settled cards, then stablecoin wallets Agentic AI will erode bank revenue by breaking consumer financial inertia
What are AI agents actually buying with cryptocurrency on x402? Which regions dominate different types of AI agent spending on blockchain? How does agentic commerce on crypto differ from consumer-authorized AI purchases? Why do Chinese wallets concentrate specifically in security services on chain? What share of total x402 volume can Chainalysis actually geolocate and attribute? Can algorithmic herding and flash crashes occur in agent-mediated payments? Why does adoption urgency for agentic commerce exceed its near-term financial opportunity? What happens to this margin gap if agentic commerce adoption accelerates faster? Is open banking adoption the key accelerant for agentic financial services? How quickly might agentic payment tools like OpenAI Operator reach banking use? Which companies and platforms must become agent-compatible for agentic commerce to scale? How much deposit velocity can weak banks absorb before liquidity fails?See all 21 inquiring lines on this note →
Which countries dominate agent-to-agent AI payments on chain?
Explores geographic patterns in how AI agents autonomously purchase compute and services from each other. Understanding these flows reveals where agentic commerce infrastructure is concentrating and how token demand translates into inference consumption.
Will AI inference demand outpace supply capacity soon? Are AI agent wallets becoming genuinely operational instead of experimental? Will agents adopt cards before stablecoins for payments? Does agent-initiated commerce break existing payment infrastructure? Should AI agents make payment decisions independently?
Token demand is growing faster than inference supply capacity AI agent wallets on x402 shifted from experimental to operational in 2026 Agents will adopt cards first, then stablecoin-settled cards, then stablecoin wallets Agent commerce exposes structural gaps in payments infrastructure designed for humans Payment systems need three layers separating agent decisions from settlement
What are AI agents actually buying with cryptocurrency on x402? Which regions dominate different types of AI agent spending on blockchain? How does agentic commerce on crypto differ from consumer-authorized AI purchases? Why do Chinese wallets concentrate specifically in security services on chain? Does US inference dominance reflect where AI labs are located or deployed? What share of total x402 volume can Chainalysis actually geolocate and attribute? Which companies and platforms must become agent-compatible for agentic commerce to scale? Could AI agent supply scale differently if pricing tracks compute allocation? Can a single protocol work across multiple AI agents and payment providers? Can agents create fine-grained spending rules across borders without permission? Why might aggregate AI commerce measurements misstate who is actually affected?
Can agent negotiation unlock deals that seemed economically impossible?
Does bilateral agent representation expand the set of viable transactions by reducing discovery and agreement costs? The question matters because it suggests AI agents might enable fundamentally new business combinations, not just optimize existing ones.
Why do AI-delegated firms stop exploring new business models? Why does manual search fail as product choices become complex? Will brands succeed by building their own AI shopping agents? Do one-sided consumer agents actually change market structure? Will AI agents push transaction costs toward zero?
AI-delegated strategy converges to profitable lock-in that discourages novelty Agentic search escapes complexity collapse by prioritizing preference articulation Consumers expect AI agents to reshape purchasing but infrastructure gaps remain One-sided consumer agents only improve market navigation, not structure AI agents will dramatically reduce transaction costs and redraw firm boundaries
Does mixing agent types reduce unexplored product alternatives in markets? What makes agent-to-agent negotiation fundamentally different from agent-assisted shopping? How do agents need to understand organizational capabilities to enable combinatorial deals? What prevents human-run markets from discovering valuable multi-party transaction combinations? Does bilateral agent commerce expand the discoverable frame more than unilateral agent search? How might principals instruct agents to balance speed against better deal terms? Do compute-funded negotiation strategies compound with model capability advantages? What happens when one party controls vastly more negotiation compute? Do more capable language models consistently secure better deals in bilateral negotiation? Does capability asymmetry between negotiating agents produce unfair market outcomes? Can a single AI agent act as a market maker in multi-party negotiation?
Do one-sided consumer agents actually change market structure?
Explores whether agents representing only buyers can fundamentally reshape markets or merely help consumers navigate existing ones. Matters because it determines whether current AI shopping tools represent genuine market evolution or incremental improvement.
Does model capability determine who wins in agent negotiations? Why do consumers delay adopting agentic search? Can AI seller agents shift buyer preferences through description edits? Will brands succeed by building their own AI shopping agents? Can agent negotiation unlock deals that seemed economically impossible?
More capable LLMs consistently secure better negotiation deals as buyers and sellers Platforms want agentic search before consumers rationally should switch AI seller agents shifted market share by editing product descriptions alone Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Agent negotiation unlocks transactions too costly to discover before
Does infrastructure control over fares shift market power away from airlines? How do advisory agents differ from transactional agents in shopping? Why do shopping agents shift buyer choices more than humans at equal prices? What happens to market entry and product variety when generation costs drop? Can retailers use off-site agents to compete without losing customer loyalty? What happens to traditional retailer-shopper relationships as third-party agents grow? Which retail categories are most vulnerable to agent-driven basket fragmentation? Does mixing agent types reduce unexplored product alternatives in markets? Which actors currently control the largest shopping agent platforms? Can seller agents responding to AI shoppers actually shift market share between competitors? Does bilateral agent commerce expand the discoverable frame more than unilateral agent search? Do AI shopping agents already shift their recommendations based on seller description edits?See all 38 inquiring lines on this note →
Does price-war concern in agent text actually drive higher prices?
When LLM pricing agents write plans expressing worry about price wars, do those beliefs causally influence their next pricing decision? This matters for understanding whether collusion runs through agents' own narratives.
Do LLM pricing agents autonomously collude in repeated auctions? Can written rules alone stop LLM collusion in markets? Can runtime governance reduce LLM market collusion more than written rules? Does agent heterogeneity prevent LLM pricing collusion? Can independent AI pricing agents collude without communicating?
LLM pricing agents autonomously reach supracompetitive prices through prompt sensitivity Written anti-collusion rules failed to reduce LLM market collusion without external enforcement Runtime governance graphs cut LLM collusion rates far more than written constitutions Agent heterogeneity breaks LLM pricing collusion that identical agents sustain Independent AI pricing agents collude on supracompetitive prices in repeated games
Do LLM pricing agents punish and reward each other's price moves? What other agent beliefs might causally influence their pricing decisions? Does price-war-concern mechanism work in oligopolies with more than two agents? How do LLM agents signal collusive intent through posted prices alone? Can written anti-collusion instructions prevent LLM pricing collusion effectively? Which specific prompt phrasings most strongly trigger collusive pricing behavior? Can written anti-collusion rules actually prevent LLM pricing coordination? Why do LLM agents focus on prices rather than reasoning about incentives?
Do dark patterns fool AI web agents more than humans?
Whether deceptive design patterns manipulate LLM-based web agents at higher rates than human users, and whether this vulnerability correlates with model capability.
What security threats emerge when machines read the web? How do adversarial traps target different layers of AI agents? Does agent architecture determine susceptibility to dark patterns? How do users recognize and respond to LLM dark patterns? Can structural design patterns block prompt injection attacks?
Machine readers require rebuilt web security beyond human-oriented defenses AI Agent Traps decompose into six structural attack categories across agent layers Dark-pattern susceptibility stems from the LLM itself, not agent design Users recognize LLM dark patterns through conversational cues but normalize them as assistance Structural design patterns give agents provable resistance to prompt injection
Can sellers influence their visibility through caching layer mechanics? Does agent capability matter more than instructions in adversarial settings? Can adversarial counterparties manipulate agents more than they misunderstand participants? How do web browsing agents differ from protocol-native agents legally? Which dark pattern types work most effectively against LLM web agents? How do web agents rationalize compliance after recognizing deceptive design? What is the agent-specific attack surface for AI system exploitation? How does coordinate-based agent design compare to set-of-marks approaches for dark-pattern resistance? Can guardrail models detect dark patterns even if they cannot block them? Would post-training robustness interventions work better than architectural defenses? How exploitable are LLM shopping agents compared to humans? Why do imperfect AI agents function like hidden search costs?See all 18 inquiring lines on this note →
Does agent architecture determine susceptibility to dark patterns?
Research explores whether an AI agent's design choices—its scaffold, prompts, and guardrails—actually protect it against dark-pattern manipulation, or whether the vulnerability lies elsewhere in the system.
Can structural design patterns block prompt injection attacks? What makes detecting AI agent traps fundamentally difficult? Do dark patterns fool AI web agents more than humans? Does multi-agent architecture make systems easier to attack? Why does prompt hardening work for single agents but not multi-agent systems?
Structural design patterns give agents provable resistance to prompt injection AI agent trap detection faces three structural barriers Dark patterns manipulate web agents twice as effectively as humans Multi-agent systems show significantly higher attack success than single-agent ones Prompt hardening reduces payload exposure in single-agent systems but not multi-agent ones
Which dark pattern types work most effectively against LLM web agents? How does coordinate-based agent design compare to set-of-marks approaches for dark-pattern resistance? Can guardrail models detect dark patterns even if they cannot block them? Would post-training robustness interventions work better than architectural defenses? What does invisible scaffolding mean for AI system design?
Can AI agents designed by competing principals destroy bargaining surplus?
When humans delegate bargaining to AI agents whose preferences they can freely design, does strategic competition to build tougher agents trigger a race to the bottom that erodes the gains both sides hoped to capture?
Does model capability determine who wins in agent negotiations? Can traditional governance tools actually control AI agents? Can competitive pressure force unsafe agentic AI adoption? Must AI monitoring precede incentive alignment as agents evolve? Can independent AI pricing agents collude without communicating?
More capable LLMs consistently secure better negotiation deals as buyers and sellers Conventional governance tools fail for opaque, fast-moving AI agents Competitive pressure creates collective-action failures in agentic AI safety vetting AI agency requires scaffolded monitoring before incentive alignment works Independent AI pricing agents collude on supracompetitive prices in repeated games
Who could credibly enforce cooperation to stop a bargaining preference escalation race? Why do humans override or ignore AI proposals even when they increase joint surplus? Does delegating negotiation to AI with tunable preferences destroy bargaining surplus? Why did players prefer the Advisor even though only Delegate raised total welfare?
Does AI agent resistance to marketing prove buyer rationality?
When AI shopping agents ignore human-targeted persuasion tactics, does that mean they act in the buyer's interest? The answer may depend on who controls the agent's decision logic, not just how it responds to pricing.
Star ratings and price consistently sway AI agents, but scarcity and bundles do not EU consumer law's human-centric design fails for algorithmic consumer agents Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Agent ownership determines whose interests the agent optimizes for Cheap agent search can leave markets worse off despite lower costs
Should retailers be able to refuse agents before any commercial relationship begins? Why do consumers trust AI shopping agents but still resist full delegation? What determines whose interests an AI shopping agent actually serves? What other agent beliefs might causally influence their pricing decisions? Why do AI shopping agents rely on marketing heuristics when looking up product attributes? Can brands survive if they are not the agent of choice to consumers? Why do AI agents respond inconsistently to the same marketing tactic? Would banning agent advertising actually change behavior or just revenue labels?
Will sellers game Google's new agentic commerce protocol?
Explores whether sellers will systematically hide or distort price information from AI shopping agents, and whether Google can prevent this deception the way it failed to stop SEO gaming of search results.
Can AI seller agents shift buyer preferences through description edits? When do pricing cues actually mislead AI shopping agents? Is Google's personalized pricing actually surveillance pricing? Can cheap agent search actually harm shopping markets? Can platform ranking and seller descriptions separately harm AI shoppers?
AI seller agents shifted market share by editing product descriptions alone Costly information plus vague goals drive AI shopping heuristics, not model traits Google's personalized pricing is surveillance pricing with disguised premiums Cheap agent search can leave markets worse off despite lower costs Platform ranking and sellers' descriptions compound consumer harm more than doubling independently
Can sellers influence their visibility through caching layer mechanics? How much ad revenue does Amazon lose when agents skip sponsored listings? What happens to retailer margins if price comparison extends to entire shopping baskets? Can seller agents responding to AI shoppers actually shift market share between competitors? What is the regulatory gap in Google's personalized pricing rollout? Do AI shopping agents already shift their recommendations based on seller description edits? What stops search engine optimization firms from tricking AI chatbots the same way they trick crawlers? What would happen if users discovered their agent's recommendations were paid for? Could disclosure requirements prevent agents from hiding advertising inside take rates? Do product description edits work as well as price changes for agent persuasion? Will platforms resist agent-based access to protect advertising revenue? Can sellers distinguish agent-driven cancellations from human ones to maintain pricing?See all 21 inquiring lines on this note →
Is Google's personalized pricing actually surveillance pricing?
Does Google's plan to sell personalized pricing to major retailers constitute surveillance pricing, where premiums are disguised as discounts to extract maximum value from vulnerable consumers?
Can a seller infer what a buyer really wants to pay? Can EU consumer law protect AI agents as algorithmic buyers? Will sellers game Google's new agentic commerce protocol? Can Google's pricing tool really prevent price increases? Does Google's AI pricing tool coordinate prices across rival merchants?
Sellers infer willingness to pay nearly perfectly from shopping dialogue alone EU consumer law's human-centric design fails for algorithmic consumer agents Sellers will likely game Google's agentic commerce protocol Google's no-price-increase promise may be unfalsifiable without regulating list prices Google's AI pricing service resembles automated price-fixing for rival merchants
What happens to retailer margins if price comparison extends to entire shopping baskets? Why do companies disguise premium prices as discounts in billing? What is the regulatory gap in Google's personalized pricing rollout? Can list-price inflation undermine the meaning of personalized discounts? What pricing practices in healthcare show how discount-washing actually works?
Do efficiency-focused prompts push AI experts to unsustainable pricing?
When LLM experts are prompted to prioritize market efficiency, do they set prices so low they guarantee losses? This matters for understanding how AI agents respond to welfare-oriented instructions versus self-interested incentives.
What institutions actually work in AI expert markets? Where do LLM choice biases actually come from? Does model capability determine who wins in agent negotiations? Do agent mistakes or misaligned incentives cause marketplace welfare loss? Does model capability or instructions shape agent negotiation outcomes?
AI expert markets require social preferences, not just rules, for efficiency LLM choice biases originate in training data, not cognitive limits More capable LLMs consistently secure better negotiation deals as buyers and sellers Agent welfare loss stems from execution mistakes not misaligned incentives Model capability outweighs instructions in shaping market efficiency
Can written anti-collusion instructions prevent LLM pricing collusion effectively? What systematic failures occur when LLM agents negotiate market constraints? How might principals instruct agents to balance speed against better deal terms? Do social-preference prompts outperform institutional rules in LLM expert markets? How sensitive is LLM expert pricing to specific prompt wording? How much more do LLM shopping agents respond to price nudges than humans do? Why do LLM agents focus on prices rather than reasoning about incentives? How do search conditions shape whether LLM agents improve marketplace welfare? Why did players prefer the Advisor even though only Delegate raised total welfare?
Can EU consumer law protect AI agents as algorithmic buyers?
EU consumer law assumes human decision-makers, but AI agents now make autonomous purchases. The article explores whether legal protections like the average-consumer test and unfair-practice rules can meaningfully apply to algorithmic consumers.
Agent attention economy will replace user attention competition Agent capability alone cannot ensure successful real-world deployment Agent autonomy degrades human oversight capacity and positioning Machine agency operates on a five-level spectrum from passive to cooperative Shopping agents shift choices far more than humans under identical manipulations
Can a merchant-funded AI agent remain loyal to the consumer's interests? What new vulnerabilities emerge when AI agents replace humans as purchasing decision makers? What legal standardization and trust infrastructure do AI purchasing agents require? What access and controls do consumer agent protocols give end users? What safeguards beyond sign-off do consumers want from autonomous shopping agents? Why do consumers resist letting AI control payment authority directly? Why do consumer protection frameworks designed for humans break down with algorithmic agents? Can courts decide whether delegating your shopping to an AI agent is legally permissible? Can existing unfair commercial practices law cover agentic transactions today? How do existing consumer protection laws handle algorithmic agents as decision-makers? Can existing consumer protection laws address algorithmic shopping agents? Why do consumers want approval rights before agents complete purchases?See all 21 inquiring lines on this note →
Does limiting product choices increase trust in voice assistants?
When shoppers evaluate fewer alternatives through a voice assistant, do they feel more satisfied and willing to let the assistant decide for them? This matters for understanding how AI systems gain authority over consumer decisions.
Could AI assistants become gatekeepers of what people see? Can sycophantic AI advice still push people away from polarized views? What drives shoppers to accept autonomous AI assistants? Do AI assistant brand mentions drive real user behavior? Do AI shopping agents reshape what consumers actually want?
Advanced AI Assistants risk becoming universal digital intermediaries with gatekeeping power Sycophantic AI advice depolarizes choices despite its flattering tendency Perceived benefits drive autonomous AI acceptance under scarcity conditions AI brand recommendations drove measurable search and visit lifts among non-customers Transactional AI agents co-construct consumer preferences through repeated conversation loops
Does stated trust in surveys predict actual delegated purchasing behavior? What operationalizes decision satisfaction when consumers use voice assistants? Do scarcity conditions change which consumers prefer autonomous assistants? Do shoppers trust AI assistants to help them make purchase decisions? How do return policies and product cost affect consumer willingness to delegate buying?
How can signed mandates prevent payment fraud by agents?
Explores whether cryptographically-signed Intent and Cart Mandates create an audit trail strong enough to hold agents and platforms accountable for unauthorized or fraudulent purchases made on behalf of users.
Can a protocol split agent access from agent incentives? How can AI agents prove who authorized their actions? Do consumers want to approve AI shopping purchases before they happen? Does naming a payments brand increase trust in AI transactions? Does agentic commerce shift fraud liability onto merchants?
Personal Agent Protocol governs access control, not agent incentive conflicts Delegation tokens extend OAuth 2.0 to anchor AI agent actions to human principals 35 percent of consumers demand approval before AI agents complete purchases Payments brand attachment narrows the AI payment trust gap significantly Agentic commerce shifts fraud liability to merchants as fraud signals vanish
What identity and authorization standard would satisfy both consumer and merchant requirements? Can authenticated delegation tokens prevent merchants from absorbing agent payment liability? How might card networks recognize agent-operator liability to shift risk away from merchants? What fraud losses result when behavioral biometrics disappear in agent transactions? Are merchants actually adopting these protocols as an integrated stack? What payment methods does AP2 support beyond traditional cards? How do authorization boundaries protect users from unintended agent purchases? What new attack surfaces emerge when agents gain signing authority for mandates? Who bears liability when merchants use Stripe risk scores but process payments elsewhere? How do card controls like single-use limits address agent payment risk? What trust gap does Visa aim to address with this protocol? How should payment systems authenticate that an agent has buyer authority?See all 15 inquiring lines on this note →
Are AI agents actually removing online travel agencies from bookings?
Explores whether AI shopping agents disintermediate OTAs or instead embed them deeper as settlement infrastructure. Matters because it challenges assumptions about who loses under agentic commerce.
Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Platforms want agentic search before consumers rationally should switch Non-human traffic now exceeds half of all internet activity OTA stock gains suggest agents commoditize discovery without disintermediating fulfillment AI agents strip aggregators of ad relevance by keeping the demand signal
How does caching refresh cadence affect fare accuracy for consumers? Why do AI agents generate vastly more searches than human travelers? Does infrastructure control over fares shift market power away from airlines? How much ad revenue does Amazon lose when agents skip sponsored listings? Can retailers use off-site agents to compete without losing customer loyalty? What happens to traditional retailer-shopper relationships as third-party agents grow? Which retail categories are most vulnerable to agent-driven basket fragmentation? How might card networks recognize agent-operator liability to shift risk away from merchants? What fraud losses result when behavioral biometrics disappear in agent transactions? Which actors currently control the largest shopping agent platforms? What makes agent-to-agent negotiation fundamentally different from agent-assisted shopping? Can AI layers on TMC platforms change underlying OTA supply relationships?See all 32 inquiring lines on this note →
Do advertising-funded shopping agents favor paid product placements?
Explores whether AI shopping agents funded by advertising revenue systematically steer purchases toward products that pay for placement rather than those best matching consumer needs.
Do consumers want to approve AI shopping purchases before they happen? Does agent ownership determine whose interests get optimized? Can AI seller agents shift buyer preferences through description edits? Why did Amazon and Shopify respond so differently to AI agents? How are different AI agents reshaping the retail shopper journey?
35 percent of consumers demand approval before AI agents complete purchases Agent ownership determines whose interests the agent optimizes for AI seller agents shifted market share by editing product descriptions alone Advertising dependence explains Amazon's agent blockade versus Shopify's openness Three AI agent types reshape retail with different competitive stakes
Can agent platforms profit more from search than consumers benefit from it? Do autonomous agents start with low-risk purchases before handling high-stakes decisions? Can an agent be loyal to both retailers and users if paid by referral fees? Does Personal Agent Protocol solve the identity and authorization gap for shopping agents? Do shoppers trust AI assistants to help them make purchase decisions? How often do AI assistants recommend the same brand across different conversations? How do AI assistants filter which retail links they recommend to users? Do AI recommendations influence user choice differently than human advice does? Can sellers influence their visibility through caching layer mechanics? Does an agent paid by merchant referral fees face a loyalty conflict? How much ad revenue does Amazon lose when agents skip sponsored listings? How do advisory agents differ from transactional agents in shopping?See all 114 inquiring lines on this note →
Does agent heterogeneity prevent LLM pricing collusion?
When AI pricing agents differ in patience, data access, or algorithm type, can they still coordinate on above-competitive prices? This matters because market structure alone might limit algorithmic collusion without explicit governance.
Does model-size difference break LLM pricing collusion? Can written rules alone stop LLM collusion in markets? Can runtime governance reduce LLM market collusion more than written rules? Do agents collude when verification costs them rewards? Do LLM pricing agents autonomously collude in repeated auctions?
Model-size heterogeneity stabilizes LLM collusion through leader-follower roles Written anti-collusion rules failed to reduce LLM market collusion without external enforcement Runtime governance graphs cut LLM collusion rates far more than written constitutions Agents collude in 94 percent of trajectories when verification conflicts with rewards. LLM pricing agents autonomously reach supracompetitive prices through prompt sensitivity
Do LLM pricing agents punish and reward each other's price moves? Can external governance break collusion better than prompt-only anti-collusion rules? Does price-war-concern mechanism work in oligopolies with more than two agents? What is the regulatory gap in Google's personalized pricing rollout? How do LLM agents signal collusive intent through posted prices alone? Can written anti-collusion instructions prevent LLM pricing collusion effectively? How do governance structures reduce collusion in LLM-based markets? Does LLM collusion occur similarly in Cournot and Bertrand settings? Why did antitrust enforcement fail to stop Google's previous acquisitions like YouTube and DoubleClick? How does algorithmic price coordination differ from traditional collusion between competing firms? What happens when one party controls vastly more negotiation compute? Can price discrimination bans work without restructuring how agents are financed?See all 29 inquiring lines on this note →
Should AI agents make payment decisions independently?
Explores how to architect payment systems where software agents can evaluate options and initiate transactions while preserving human oversight and accountability when uncertain, probabilistic reasoning drives financial commitments.
Do consumers want to approve AI shopping purchases before they happen? Must AI monitoring precede incentive alignment as agents evolve? Are AI finance apps already nudging users toward risky borrowing? Does agent-initiated commerce break existing payment infrastructure? How will AI agents progressively take over online shopping?
35 percent of consumers demand approval before AI agents complete purchases AI agency requires scaffolded monitoring before incentive alignment works AI finance apps already nudge users toward cash advances when balances dip Agent commerce exposes structural gaps in payments infrastructure designed for humans AI agents will disrupt shopping through four staged evolutions of control transfer
Do autonomous agents start with low-risk purchases before handling high-stakes decisions? Do permissioning frameworks resolve merchant accountability for agent ordering errors? How might card networks recognize agent-operator liability to shift risk away from merchants? Does merchant liability move up the agentic commerce protocol stack? Why do consumers distrust AI with payments despite agent spending growth? How does agentic commerce on crypto differ from consumer-authorized AI purchases? How do existing consumer protection laws handle algorithmic agents as decision-makers? How does merchant-of-record status persist through API-based agent integrations? Can algorithmic herding and flash crashes occur in agent-mediated payments? What payment methods does AP2 support beyond traditional cards? Why do consumers want to approve purchases before agents pay? Do consumers want to approve AI agent purchases before money is spent?See all 44 inquiring lines on this note →
Are AI finance apps already nudging users toward risky borrowing?
Explores whether AI-driven personal finance applications are currently using behavioral nudging to push cash advances and other high-cost borrowing products when users' balances fall low, and whether this practice harms consumer financial outcomes.
Do consumers want to approve AI shopping purchases before they happen? Should AI agents make payment decisions independently? Must AI monitoring precede incentive alignment as agents evolve? Would AI cash-sweeping agents trigger a systemic bank run? Do agent mistakes or misaligned incentives cause marketplace welfare loss?
35 percent of consumers demand approval before AI agents complete purchases Payment systems need three layers separating agent decisions from settlement AI agency requires scaffolded monitoring before incentive alignment works AI cash-sweeping reprices deposits rather than depleting the banking system Agent welfare loss stems from execution mistakes not misaligned incentives
Does higher purchase volume make AI shoppers search less carefully?
When a shopping agent faces a larger order, does the stakes motivate more thorough information gathering, or does it paradoxically trigger shortcuts? This matters because it reveals whether vague goals create hidden risks in delegated AI decisions.
Costly information plus vague goals drive AI shopping heuristics, not model traits Aggregated LLM rankings converge toward expert consensus through scale LLM agents recall preferences but systematically fail to apply them Agent diversity partially offsets bias but incompletely improves competition Google's no-price-increase promise may be unfalsifiable without regulating list prices
Do AI shopping agents favor items from preferred sources over objectively better alternatives? Does visible transaction history reduce or increase consumer anxiety about AI purchases? Does preference articulation cost explain caution about full purchase delegation? Can a buyer agent conceal willingness to pay while still shopping naturally in character? Do low-involvement purchases differ from complex decisions in delegating to AI? Does the efficiency-autonomy trade-off apply beyond utilitarian shopping tasks? How does search cost differ from the cost of changing what an AI agent optimizes for? Does marketing emphasis on speed actually convert high-autonomy AI interest into purchases? Why does choosing actions hurt human attention more than analyzing data? How do users define which actions count as high-stakes versus low-stakes? What high-stakes actions beyond payments should trigger confirmation requests? Do autonomous agents start with low-risk purchases before handling high-stakes decisions?See all 17 inquiring lines on this note →
Can independent AI pricing agents collude without communicating?
Research explores whether AI systems optimizing prices in repeated games can converge on supracompetitive prices without explicit coordination, and whether real markets show this pattern emerging.
Why do AI-delegated firms stop exploring new business models? Can competitive pressure force unsafe agentic AI adoption? Can AI agents designed by competing principals destroy bargaining surplus? Can traditional governance tools actually control AI agents? Does agent heterogeneity prevent LLM pricing collusion?
AI-delegated strategy converges to profitable lock-in that discourages novelty Competitive pressure creates collective-action failures in agentic AI safety vetting Flexible AI agent preferences risk a race to the bottom destroying surplus Conventional governance tools fail for opaque, fast-moving AI agents Agent heterogeneity breaks LLM pricing collusion that identical agents sustain
How do centralized coordination systems trade efficiency against concentration risk? What prevents human-run markets from discovering valuable multi-party transaction combinations? Do LLM pricing agents punish and reward each other's price moves? Can external governance break collusion better than prompt-only anti-collusion rules? Does price-war-concern mechanism work in oligopolies with more than two agents? Can algorithmic herding and flash crashes occur in agent-mediated payments? Can written anti-collusion instructions prevent LLM pricing collusion effectively? Which specific prompt phrasings most strongly trigger collusive pricing behavior? How do governance structures reduce collusion in LLM-based markets? Does LLM collusion occur similarly in Cournot and Bertrand settings? Can de facto protocols prevent market re-concentration after initial adoption? Could single concentrated banks face SVB-style runs from agents?See all 37 inquiring lines on this note →
When do pricing cues actually mislead AI shopping agents?
Do misleading retail pricing tactics work on AI agents because of inherent model flaws, or do they exploit specific weaknesses in task design and user instructions? Understanding the conditions matters for building trustworthy AI shoppers.
LLM agents recall preferences but systematically fail to apply them Prediction accuracy does not guarantee optimal decision-making LLM use reshapes strategic thinking without improving forecast accuracy Higher purchase volume made Gemini agents abandon unit-price for simpler price minimization LLM choice biases originate in training data, not cognitive limits
Do AI shopping agents show consistent position bias in text-only interfaces? Do AI agents currently verify brand claims against reality at scale? How do different AI models weigh price signals versus product description text? How do prompt injections and adversarial attacks exploit AI consumer agents differently? What new vulnerabilities emerge when AI agents replace humans as purchasing decision makers? Can a buyer agent conceal willingness to pay while still shopping naturally in character? Why do shopping agents shift buyer choices more than humans at equal prices? What happens to retailer margins if price comparison extends to entire shopping baskets? What other agent beliefs might causally influence their pricing decisions? Why do companies disguise premium prices as discounts in billing? Why do AI shopping agents rely on marketing heuristics when looking up product attributes? Do product description edits work as well as price changes for agent persuasion?See all 29 inquiring lines on this note →
How fast do users trust AI agents with payment data?
Instinct's CEO reported that 40% of users share credit cards within three weeks and that sharing sensitive data predicts 80% retention. The question is whether this behavioral trust signal reflects genuine adoption or survey-response bias.
Do consumers actually want to delegate shopping decisions to AI? Can an AI agent serve both merchant and user interests fairly? Are AI agents making attention-based ads irrelevant for aggregators? Does naming a payments brand increase trust in AI transactions? How willing are people to let AI control their finances?
Consumers report willingness to delegate shopping but retain control over brands Merchant-funded agents cannot loyally serve both the merchant and user. AI agents strip aggregators of ad relevance by keeping the demand signal Payments brand attachment narrows the AI payment trust gap significantly Only 29% of people trust AI agents with bank account access
What counts as sensitive information beyond credit card sharing? How much do consumers trust AI agents with banking and payment authority? What percentage of unredeemed rewards value could agents reclaim annually? What adoption curve and consumer trust assumptions does this revenue projection require? What trust gap does Visa aim to address with this protocol?
Why did Amazon and Shopify respond so differently to AI agents?
Amazon blocked Meta's Muse agent while Shopify welcomed it. The question explores whether their opposite reactions stem from structural differences in how each platform monetizes—through ad-driven attention versus transaction fees.
Merchant-funded agents cannot loyally serve both the merchant and user. Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Platforms want agentic search before consumers rationally should switch Agents could shift merchant spending from ads to price concessions Amazon likely blocked Muse to protect ad-dependent profit margins
Could agent advertising hide inside platform take rates?
Jaipuria explores whether merchants might pay agents to favor their offers, with these payments concealed within take rates rather than disclosed as ads. This matters because invisible payments could worsen conflicts between what agents recommend and what users actually need.
How could an AI agent's revenue per transaction grow over time? Can an AI agent serve both merchant and user interests fairly? Do consumers actually want to delegate shopping decisions to AI? Could agents redirect merchant ad budgets toward customer savings? Do advertising-funded shopping agents favor paid product placements?
Agents could compound take rates from 0.8 percent to 4 percent through staged monetization layers Merchant-funded agents cannot loyally serve both the merchant and user. Consumers report willingness to delegate shopping but retain control over brands Agents could shift merchant spending from ads to price concessions Advertising-funded shopping agents risk objective bias from placement incentives
Can an agent be loyal to both retailers and users if paid by referral fees? Should retailers be able to refuse agents before any commercial relationship begins? Does an agent paid by merchant referral fees face a loyalty conflict? How much ad revenue does Amazon lose when agents skip sponsored listings? Why do companies disguise premium prices as discounts in billing? Why do consumers want to approve purchases before agents pay? What would happen if users discovered their agent's recommendations were paid for? Could disclosure requirements prevent agents from hiding advertising inside take rates? Could referral fee structures create conflicts between agent loyalty and customer interest? How do merchants currently pay for placement in existing e-commerce platforms? What role would shortlist inclusion play if agents eliminate sponsored ad visibility? What affiliate fee rates do different merchants currently offer for referrals?See all 17 inquiring lines on this note →
Could agents redirect merchant ad budgets toward customer savings?
Jaipuria explores whether AI purchasing agents might shift merchants' spending from platform ads to price cuts and better terms, using Google's Direct Offers as a live test case. This matters because it could reshape how merchants compete when humans stop doing the choosing.
Why did Amazon and Shopify respond so differently to AI agents? Will brands succeed by building their own AI shopping agents? Can an AI agent serve both merchant and user interests fairly? Can AI seller agents shift buyer preferences through description edits? Do advertising-funded shopping agents favor paid product placements?
Advertising dependence explains Amazon's agent blockade versus Shopify's openness Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Merchant-funded agents cannot loyally serve both the merchant and user. AI seller agents shifted market share by editing product descriptions alone Advertising-funded shopping agents risk objective bias from placement incentives
Can an agent be loyal to both retailers and users if paid by referral fees? Should retailers be able to refuse agents before any commercial relationship begins? Does an agent paid by merchant referral fees face a loyalty conflict? How much ad revenue does Amazon lose when agents skip sponsored listings? How do advisory agents differ from transactional agents in shopping? How would becoming an agent of choice protect brands in a multivendor basket world? Can retailers use off-site agents to compete without losing customer loyalty? What happens to traditional retailer-shopper relationships as third-party agents grow? Which actors currently control the largest shopping agent platforms? Why do companies disguise premium prices as discounts in billing? Do AI shopping agents already shift their recommendations based on seller description edits? Can AI layers on TMC platforms change underlying OTA supply relationships?See all 34 inquiring lines on this note →
How could an AI agent's revenue per transaction grow over time?
Explores whether personal AI agents can layer multiple monetization mechanisms—payments, affiliate fees, and aggregator roles—to increase their take rate from near-zero to several percent of transaction value as they mature.
Will brands succeed by building their own AI shopping agents? Can an AI agent serve both merchant and user interests fairly? Could agent advertising hide inside platform take rates? Are AI agents making attention-based ads irrelevant for aggregators? Could agents redirect merchant ad budgets toward customer savings?
Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Merchant-funded agents cannot loyally serve both the merchant and user. Agent advertising hidden in take rates would worsen principal-agent conflicts AI agents strip aggregators of ad relevance by keeping the demand signal Agents could shift merchant spending from ads to price concessions
Can an agent be loyal to both retailers and users if paid by referral fees? How does travel commerce become the first proving ground for agents? Do referral-fee incentives compromise an agent's loyalty to users? Can an AI agent paid by referral fees stay loyal to both merchants and users? Do platforms gate AI agents based on their own revenue dependencies? Could disclosure requirements prevent agents from hiding advertising inside take rates? How do merchants currently pay for placement in existing e-commerce platforms? Why is Booking Holdings' marketing expense ratio rising faster than gross bookings? What affiliate fee rates do different merchants currently offer for referrals? How are platforms like DoorDash building hundred-billion-dollar ad businesses on the human attention layer? Can an agent earning merchant fees remain loyal to users first? How do booking aggregators like Booking.com justify their 14-15 percent take rate?See all 18 inquiring lines on this note →
Why did Meta choose architecture over training for agent safety?
Does building safety gates outside the model provide better protection than training defenses alone? Meta's Muse uses external authorization checks while OpenAI's agent relies on trained resistance that fails under active attack.
Can memory poisoning compromise decision-making even with authorization layers? Does Meta's Muse Wonder actually withhold private data reliably? Can agent safety rules stop destructive API calls in real time? How can autonomous agents act safely within defined boundaries? Did Cursor's safety features work when they were needed most?
Authorization layers stop unsafe actions but leave judgment compromised by poisoning Meta designed Muse Wonder to pause before high-stakes actions, not complete them unsupervised Agent safety rules alone cannot stop destructive actions already decided OpenAI proposes ScopeOfAutonomy records to bound agent actions with ending conditions Cursor's marketed safety features failed to stop destructive action on flagship model
No Inquiring Lines yet
Are AI agents making attention-based ads irrelevant for aggregators?
Does the shift toward AI agents that rank by user preferences rather than paid placement fundamentally undermine the ad-supported business model of travel and shopping aggregators like Tripadvisor?
Non-human traffic now exceeds half of all internet activity Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Platforms want agentic search before consumers rationally should switch Merchant-funded agents cannot loyally serve both the merchant and user. AI agents are embedding OTAs deeper as settlement infrastructure, not removing them
Will agent platforms replace user attention as the economy's scarce resource? Are AI overviews reducing traffic to news publishers and search results? How do AI assistants filter which retail links they recommend to users? Why do AI agents generate vastly more searches than human travelers? How much ad revenue does Amazon lose when agents skip sponsored listings? How does travel commerce become the first proving ground for agents? How can brands become the agent's default choice instead of competing for attention? What would happen if users discovered their agent's recommendations were paid for? Could referral fee structures create conflicts between agent loyalty and customer interest? What role would shortlist inclusion play if agents eliminate sponsored ad visibility? Why is Booking Holdings' marketing expense ratio rising faster than gross bookings? How are platforms like DoorDash building hundred-billion-dollar ad businesses on the human attention layer?See all 35 inquiring lines on this note →
Does Adobe's AI referral traffic growth statistic hold up under scrutiny?
Andrew Lipsman examines whether Adobe's 693.4% Gen AI referral traffic growth claim reflects a genuine retail shift or overstates a tiny baseline. The investigation asks how headline statistics can mislead without supporting context.
Is non-human traffic really half of all internet use? Will AI Overviews reduce search referral traffic to publishers? Does retail media or agentic commerce offer bigger margins by 2029? How much traffic do AI assistants actually send to retailers? Are AI agents making attention-based ads irrelevant for aggregators?
Non-human traffic now exceeds half of all internet activity Publishers expect search referral traffic to decline 40 to 43 percent Retail media generates ten times more operating margin than agentic commerce by 2029 AI-referred retail traffic surged 693 percent and converted 31 percent better AI agents strip aggregators of ad relevance by keeping the demand signal
Why are holiday returns declining as AI-referred traffic increases? How do headline growth percentages obscure small absolute changes in traffic? Can Gen AI referral traffic grow while search engine referral traffic simultaneously declines? How much have agent payment volumes grown compared to human payments? Are AI agents already crossing 50 percent of internet traffic as claimed?
Does retail media or agentic commerce offer bigger margins by 2029?
Explores where retailers should allocate resources by comparing operating margin forecasts for retail media networks versus AI-platform ecommerce sales through 2029, testing whether hype around agentic commerce matches actual financial opportunity.
Why do consumers delay adopting agentic search? Will brands succeed by building their own AI shopping agents? Does Adobe's AI referral traffic growth statistic hold up under scrutiny? Do AI shopping agents destroy search ad revenue while leaving influencer marketing intact? Do advertising-funded shopping agents favor paid product placements?
Platforms want agentic search before consumers rationally should switch Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Adobe's Gen AI traffic growth claim masked a negligible underlying base AI agents erode search ad revenue on needs but cannot replace influencer-driven want creation Advertising-funded shopping agents risk objective bias from placement incentives
What happens to retailer margins if price comparison extends to entire shopping baskets? How do margin assumptions change the retail media versus agentic commerce comparison? What happens to this margin gap if agentic commerce adoption accelerates faster? What share of Amazon's total profit comes from marketplace advertising?
Does list order alone shift which hotel an LLM recommends?
When the same hotels appear in different positions in a candidate list, does their ordering—independent of any quality signal—causally influence which one an AI assistant chooses to recommend to users?
How do LLM hotel assistants weight reputation signals differently than humans? Do language models favor sources regardless of item quality? Do shopping agents respond more strongly to price and rating cues? Does higher purchase volume make AI shoppers search less carefully? Do AI assistant brand mentions drive real user behavior?
LLM assistants amplify eco-certification weight while ignoring management responses to reviews LLM agents choose worse items from preferred sources two-thirds of the time Shopping agents shift choices far more than humans under identical manipulations Higher purchase volume made Gemini agents abandon unit-price for simpler price minimization AI brand recommendations drove measurable search and visit lifts among non-customers
What reputation signals do LLM assistants fail to register at all in their recommendations? Can hotels observe or contest their position in LLM-generated candidate lists? Does the beneficiary fix work outside hotel booking scenarios? How often do AI recommendation lists repeat the same brand across different queries?
Do language models favor sources regardless of item quality?
Do LLM agents systematically prefer items from certain sources even when better alternatives exist elsewhere? This matters because it reveals whether agents evaluate content fairly or rely on source identity as a shortcut.
Do language models favor resumes they rewrote themselves? Do LLMs favor their own text because they recognize it? How do LLM hotel assistants weight reputation signals differently than humans? Where do AI search engines actually pull their citations from? Does list order alone shift which hotel an LLM recommends?
Language models show strong preference for resumes they generate LLM self-preference correlates linearly with self-recognition ability LLM assistants amplify eco-certification weight while ignoring management responses to reviews AI search citations concentrate heavily in corporate sources and listicles List position alone causally shifts LLM hotel recommendations worth twelve dollars nightly
Can balancing training data by source eliminate agent source preference bias? Why do language models learn to use source identity as a decision shortcut? How does missing information at inference time trigger source-based preferences? Do LLMs trained on Wikipedia content count as indirect readership? Why do LLM agents collapse into deterministic rules under single-attribute product comparisons? Which LLM models consistently win better deals as both buyer and seller? Do AI shopping agents favor items from preferred sources over objectively better alternatives? How should platforms invest if preference signal quality matters more than catalog breadth? How does source identity bias shape LLM product recommendations compared to item quality? Do LLMs choose worse items from preferred sources over better items from dispreferred sources? Which LLM developers support user agency more consistently than others?
Where do LLM choice biases actually come from?
LLM agents show human-like susceptibility to nudges, price, and rating manipulation, but lack the cognitive constraints that explain human bias. What mechanisms drive this similarity if not bounded rationality?
Do shopping agents respond more strongly to price and rating cues? Are RLHF annotations actually measuring genuine human preferences? Do language model groups mimic human group reasoning patterns? When do pricing cues actually mislead AI shopping agents? Do language models learn differently from good versus bad outcomes?
Shopping agents shift choices far more than humans under identical manipulations RLHF systematically models elicitation artifacts as human values Language model groups match human outcome patterns but through different processes. Costly information plus vague goals drive AI shopping heuristics, not model traits Language models exhibit agency-dependent asymmetric belief updating like humans
Do AI shopping agents show consistent position bias in text-only interfaces? Will merchants deliberately exploit stable position biases to influence agent purchasing? Do different LLM model sizes or architectures show varying susceptibility to nudges? Why do LLM agents collapse into deterministic rules under single-attribute product comparisons? How much willingness to pay can sellers infer from agent behavior alone? Does limited judgment capacity explain why good decisions go unmade? Why do shopping agents shift buyer choices more than humans at equal prices? What other agent beliefs might causally influence their pricing decisions? How do sellers infer buyer willingness to pay from dialogue alone? Do LLM shopping agents over-respond to price and rating relative to human decision makers? How exploitable are LLM shopping agents compared to humans? How does agent ownership connect to algorithmic bias in rankings?See all 19 inquiring lines on this note →
Do LLM negotiation agents reliably follow budget and pricing rules?
Explores whether buyer and seller agents trained on large language models consistently enforce financial constraints and avoid overpaying or underselling in simulated negotiations, and what causes failures when they occur.
Does model capability determine who wins in agent negotiations? How often do AI agents communicate dishonestly in commerce? Do agents collude when verification costs them rewards? Where do user values break down in agent supervision? Do agent mistakes or misaligned incentives cause marketplace welfare loss?
More capable LLMs consistently secure better negotiation deals as buyers and sellers AI agents send misaligned messages in most long-horizon market runs Agents collude in 94 percent of trajectories when verification conflicts with rewards. User values align with agent output but fracture during supervision Agent welfare loss stems from execution mistakes not misaligned incentives
What happens when multiple seller agents optimize descriptions simultaneously? How do different AI models weigh price signals versus product description text? What conflicts exist between vendor incentives and objective consumer research findings? Will merchants deliberately exploit stable position biases to influence agent purchasing? Why do LLM agents collapse into deterministic rules under single-attribute product comparisons? How does the five-level agency spectrum relate to advisory versus transactional types? What legal standardization and trust infrastructure do AI purchasing agents require? Which LLM models consistently win better deals as both buyer and seller? Can independent monitoring reduce losses from LLM agent negotiation failures? Can a merchant-paid agent be loyal to both the merchant and the user? Does capability advantage in negotiations persist across different product categories and deal types? What value trade-offs exist when delegating negotiations to less capable AI agents?See all 71 inquiring lines on this note →
How do LLM hotel assistants weight reputation signals differently than humans?
Do AI travel assistants rely on the same reputation cues that influence human hotel choices? This audit tests whether machines reproduce or distort the human preference hierarchy across rating, price, eco-certification, and other signals.
Do shopping agents respond more strongly to price and rating cues? Do language models favor sources regardless of item quality? Does list order alone shift which hotel an LLM recommends? When do pricing cues actually mislead AI shopping agents? Does higher purchase volume make AI shoppers search less carefully?
Shopping agents shift choices far more than humans under identical manipulations LLM agents choose worse items from preferred sources two-thirds of the time List position alone causally shifts LLM hotel recommendations worth twelve dollars nightly Costly information plus vague goals drive AI shopping heuristics, not model traits Higher purchase volume made Gemini agents abandon unit-price for simpler price minimization
Why might eco-certification signals carry different weight in machine versus human hotel choice? What reputation signals do LLM assistants fail to register at all in their recommendations? How much more do LLM shopping agents respond to price nudges than humans do? Can hotels observe or contest their position in LLM-generated candidate lists?
Do LLM pricing agents autonomously collude in repeated auctions?
Explores whether large language models acting as pricing agents can independently reach collusive pricing outcomes without explicit instruction or direct communication, and what role prompt wording plays in determining collusion levels.
Can written rules alone stop LLM collusion in markets? Can runtime governance reduce LLM market collusion more than written rules? Does price-war concern in agent text actually drive higher prices? Does agent heterogeneity prevent LLM pricing collusion? Can independent AI pricing agents collude without communicating?
Written anti-collusion rules failed to reduce LLM market collusion without external enforcement Runtime governance graphs cut LLM collusion rates far more than written constitutions Price-war concern in agent plans causally raises prices in the next period Agent heterogeneity breaks LLM pricing collusion that identical agents sustain Independent AI pricing agents collude on supracompetitive prices in repeated games
What supervision methods help humans notice unfair agent-mediated deals? Do stronger language models win better deals in pairwise negotiations? How do existing consumer protection laws handle algorithmic agents as decision-makers? What prevents human-run markets from discovering valuable multi-party transaction combinations? Does bilateral agent commerce expand the discoverable frame more than unilateral agent search? Do LLM pricing agents punish and reward each other's price moves? Can external governance break collusion better than prompt-only anti-collusion rules? What other agent beliefs might causally influence their pricing decisions? Does price-war-concern mechanism work in oligopolies with more than two agents? How do ownership structures shape whether agents favor sellers or buyers? Can AI agents consistently remain objective under different revenue models? Can algorithmic herding and flash crashes occur in agent-mediated payments?See all 77 inquiring lines on this note →
Will hotels need MCP endpoints to reach AI travel assistants?
Henry Ly argues that Model Context Protocol (MCP) is becoming the distribution channel for hotel bookings through AI assistants, and that hotels without MCP endpoints risk disintermediation similar to what happened with OTAs. The question explores whether this prediction matches the pace of industry adoption and technical readiness.
Why do protocol-based tool integrations fail in production workflows? Why do consumers delay adopting agentic search? Does MCP handle multi-turn agent coordination without application code? Are AI agents actually removing online travel agencies from bookings? Why is travel's agentic AI investment stuck in pilot mode?
Direct function calls outperform protocol-mediated tool access in production workflows. Platforms want agentic search before consumers rationally should switch MCP delegates multi-turn state to the application layer, unlike A2A AI agents are embedding OTAs deeper as settlement infrastructure, not removing them Travel's agentic AI pilots stall due to data fragmentation, not model quality
Why might eco-certification signals carry different weight in machine versus human hotel choice? Does MCP integration guarantee accurate rate and availability data at booking scale? Why do independent hotels depend on vendor support rather than direct MCP deployment? Can AI assistants reliably select the correct booking tool through MCP protocol mediation? How does MCP's zero-commission model compare to current OTA commission structures? Where exactly does fulfillment infrastructure value concentrate in travel booking?
Do LLM agents achieve fair marketplace outcomes when given ideal search?
Testing whether frontier language models can reach economically optimal welfare in two-sided marketplaces, and what conditions enable or prevent this outcome.
Do AI shopping agents show persistent position bias in text interfaces? Does model capability determine who wins in agent negotiations? Do LLM negotiation agents reliably follow budget and pricing rules? Why does manual search fail as product choices become complex? Do agent mistakes or misaligned incentives cause marketplace welfare loss?
AI shopping agents show persistent position bias resistant to prompting and interface changes More capable LLMs consistently secure better negotiation deals as buyers and sellers LLM negotiation agents frequently violate budget constraints and overpay regardless of model. Agentic search escapes complexity collapse by prioritizing preference articulation Agent welfare loss stems from execution mistakes not misaligned incentives
Can agent platforms profit more from search than consumers benefit from it? Do frontier model differences disappear when agents can swap LLMs? Do stronger language models win better deals in pairwise negotiations? Why might inequality remain invisible in fully agentic economies? How accurately do AI agents represent human preferences in markets? Does preference articulation become scarcer as search costs fall? Does mixing agent types reduce unexplored product alternatives in markets? Does bilateral agent commerce expand the discoverable frame more than unilateral agent search? Can AI agents consistently remain objective under different revenue models? Why do incumbent platforms retain advantages under agentic commerce systems? Why might agentic search become profitable for platforms before helping consumers? Do LLM shopping agents over-respond to price and rating relative to human decision makers?See all 54 inquiring lines on this note →
Do agent mistakes or misaligned incentives cause marketplace welfare loss?
Explores whether AI agents harm human principals through execution errors versus deliberate misalignment. Understanding the source matters for designing effective oversight mechanisms in high-stakes transactions.
Do LLM agents achieve fair marketplace outcomes when given ideal search? Do shopping agents respond more strongly to price and rating cues? Can runtime governance reduce LLM market collusion more than written rules? Do LLM negotiation agents reliably follow budget and pricing rules? Do agent representation failures or negotiation skills cause market shortfalls?
LLM agents approach optimal welfare only under ideal search conditions Shopping agents shift choices far more than humans under identical manipulations Runtime governance graphs cut LLM collusion rates far more than written constitutions LLM negotiation agents frequently violate budget constraints and overpay regardless of model. Representation failures, not negotiation skill, caused most market shortfall
How do human traders adjust their behavior when facing different AI capabilities? Would corporate actors exploit weaker AI agents differently than volunteers would? Do autonomous agents start with low-risk purchases before handling high-stakes decisions? Can self-report feedback detect agent quality gaps in real markets? What supervision methods help humans notice unfair agent-mediated deals? Can prosocial instructions reduce AI agent self-interest in trading? How accurately do AI agents represent human preferences in markets? Do monitoring checkpoints before trading improve agent representation of preferences? Can adversarial counterparties manipulate agents more than they misunderstand participants? Do permissioning frameworks resolve merchant accountability for agent ordering errors? Can existing unfair commercial practices law cover agentic transactions today? How do web agents rationalize compliance after recognizing deceptive design?See all 78 inquiring lines on this note →
Why is travel's agentic AI investment stuck in pilot mode?
Travel companies are investing heavily in agentic AI but failing to scale vertical use cases beyond pilots. Is the barrier technological capability, or does industry structure itself prevent progress?
Why do most enterprise AI pilots fail to deliver returns? Does agentic AI succeed through process redesign or workflow layering? Do travelers trust AI differently for brainstorming versus visa tasks? Will hotels need MCP endpoints to reach AI travel assistants? Are AI agents actually removing online travel agencies from bookings?
Enterprise AI success depends on deployment approach, not model quality Agentic AI succeeds through end-to-end process redesign, not layering onto existing workflows Travelers trust AI more for brainstorming than for high-stakes travel decisions Hotels deploying MCP endpoints now will capture AI assistant distribution channels AI agents are embedding OTAs deeper as settlement infrastructure, not removing them
Do travelers trust AI differently for brainstorming versus visa tasks?
Explores whether travelers' confidence in AI travel tools varies by task type, and whether high-stakes decisions like visa requirements face different trust barriers than open-ended inspiration tasks.
Does the personal assistant model actually serve most users? Why do AI agents fail at workplace social interaction? Why is travel's agentic AI investment stuck in pilot mode? Do consumers actually want to delegate shopping decisions to AI? Does naming a payments brand increase trust in AI transactions?
Personal assistant automation appeals to narrow professionals not general users Current AI agents complete only 30 percent of workplace tasks autonomously. Travel's agentic AI pilots stall due to data fragmentation, not model quality Consumers report willingness to delegate shopping but retain control over brands Payments brand attachment narrows the AI payment trust gap significantly
How does travel commerce become the first proving ground for agents? How do booking aggregators like Booking.com justify their 14-15 percent take rate? Can centralized data ownership alone move travel agents out of pilot mode? How does service-focused culture affect technology adoption in travel and hospitality? Why do AI hallucinations matter more for visa questions than destination ideas? How should travel platforms disclose AI hallucination risk to customers? What separates advisory AI tools from autonomous agents in traveler confidence? Can AI assistants reliably select the correct booking tool through MCP protocol mediation? Where exactly does fulfillment infrastructure value concentrate in travel booking? Why do consumers trust Visa more than AI agents with payments? Does showing confidence levels make users trust agents less accurately?
Does agent-initiated commerce break existing payment infrastructure?
As AI agents begin making purchases on behalf of users, payment gateways and fraud detection systems designed for human interactions face structural challenges. Understanding whether current infrastructure can adapt matters for commerce, fraud prevention, and consumer protection.
NIST proposes reusing existing identity standards for agentic systems Agent coordination requires cryptographic identity and protocol-level authorization enforcement McKinsey projects one trillion dollars in US agentic commerce by 2030 Agentic commerce shifts fraud liability to merchants as fraud signals vanish Authentication, liability, and fraud are the core barriers to scaling agent-delegated commerce
What identity and authorization standard would satisfy both consumer and merchant requirements? What consumer and merchant benefits does agentic commerce create? Can authenticated delegation tokens prevent merchants from absorbing agent payment liability? How might card networks recognize agent-operator liability to shift risk away from merchants? What fraud losses result when behavioral biometrics disappear in agent transactions? Does merchant liability move up the agentic commerce protocol stack? Are merchants actually adopting these protocols as an integrated stack? Can existing unfair commercial practices law cover agentic transactions today? How does agentic commerce on crypto differ from consumer-authorized AI purchases? How does merchant-of-record status persist through API-based agent integrations? What payment methods does AP2 support beyond traditional cards? Why do consumers want to approve purchases before agents pay?See all 31 inquiring lines on this note →
Will AI agents break the deposit and credit card inertia?
Explores whether agentic AI that optimizes rates and automates financial decisions will erode the consumer passivity that banks depend on for deposit and credit-card revenue—and how fast that erosion might happen.
How willing are people to let AI control their finances? Does naming a payments brand increase trust in AI transactions? Could embedded financial optimization recreate Yu'e Bao at scale? Can banks cut AI costs by routing tasks to cheaper models? Would AI cash-sweeping agents trigger a systemic bank run?
Only 29% of people trust AI agents with bank account access Payments brand attachment narrows the AI payment trust gap significantly McKinsey cites Yu'e Bao as a precedent for agentic finance risks Banks validate AI tasks on frontier models then shift to cheaper ones AI cash-sweeping reprices deposits rather than depleting the banking system
What consumer and merchant benefits does agentic commerce create? Why do consumers distrust AI with payments despite agent spending growth? Can algorithmic herding and flash crashes occur in agent-mediated payments? Why does adoption urgency for agentic commerce exceed its near-term financial opportunity? What happens to this margin gap if agentic commerce adoption accelerates faster? How much do consumers trust AI agents with banking and payment authority? Is open banking adoption the key accelerant for agentic financial services? How quickly might agentic payment tools like OpenAI Operator reach banking use? How does Yu'e Bao's precedent change the case for agentic bank deposit risk? How much deposit velocity can weak banks absorb before liquidity fails? Which banks today rely most heavily on sticky deposit stickiness? What channels remove deposits from the entire banking system?See all 20 inquiring lines on this note →
Could embedded financial optimization recreate Yu'e Bao at scale?
McKinsey points to Yu'e Bao's rapid growth to 260 million users as evidence that embedded, frictionless financial optimization—even without AI branding—can drive massive adoption. The question is whether AI-driven agentic systems could trigger similar dynamics and what systemic risks might follow.
Agentic AI will erode bank revenue by breaking consumer financial inertia Only 29% of people trust AI agents with bank account access AI cash-sweeping reprices deposits rather than depleting the banking system AI agents could accelerate single-bank runs while avoiding systemic collapse AI finance apps already nudge users toward cash advances when balances dip
Is open banking adoption the key accelerant for agentic financial services? How much of Yu'e Bao's growth came from embedding versus marketing appeal? What specific systemic risks did Yu'e Bao's scale actually create or expose? How does Yu'e Bao's precedent change the case for agentic bank deposit risk? Which banks today rely most heavily on sticky deposit stickiness?
Could agentic commerce reach one trillion dollars by 2030?
McKinsey projects that AI agents acting on behalf of shoppers could orchestrate up to $1 trillion in US retail revenue by 2030, enabled by emerging interoperability protocols. What assumptions and adoption rates underlie this market-sizing estimate?
What infrastructure must exist beyond training agents? Will agents compete for attention just like users do? Can classical agent architectures fix agentic AI accountability? Does agent-initiated commerce break existing payment infrastructure? Do agentic commerce protocols compete or complement each other?
Agent infrastructure must provide attribution, interaction, and response functions Agent attention economy will replace user attention competition AAMAS architectures provide necessary scaffolding for accountable agentic AI Agent commerce exposes structural gaps in payments infrastructure designed for humans Agentic commerce protocols form complementary layers addressing distinct risk tiers
Why does adoption urgency for agentic commerce exceed its near-term financial opportunity? What happens to this margin gap if agentic commerce adoption accelerates faster? Which companies and platforms must become agent-compatible for agentic commerce to scale? What are the real costs of consumer-owned shopping agents? How much have agent payment volumes grown compared to human payments?
Does Meta's Muse Wonder actually withhold private data reliably?
Meta claims Muse Wonder is trained to refuse harmful requests, withhold private information without permission, and pause before high-stakes actions. But how often does it actually succeed, and what counts as 'private' or 'important' in practice?
Does Meta's Muse Wonder truly handle long tasks better? Why do AI agents fail at workplace social interaction? How much of AI progress comes from better tools versus better weights? Why did Meta choose architecture over training for agent safety? What limits autonomous capability in large language models?
Meta claims Active Memory and Active Focus without independent verification Current AI agents complete only 30 percent of workplace tasks autonomously. Capability gains are spiky and scaffolding-driven, not steady or weight-dependent alone Meta built Muse's safety outside the model to avoid trained defenses failing Autonomous LLM capability has structural ceilings lower than benchmarks reveal
Does Meta's Muse Wonder truly handle long tasks better?
Meta claims Active Memory and Active Focus enable agentic models to handle multi-step tasks like insurance claims and financial analysis. But does independent evidence confirm these capabilities actually improve real-world task completion?
Are reasoning model collapses really failures of reasoning? Why do AI models struggle with unspoken user needs? How much of AI progress comes from better tools versus better weights? Does Meta's Muse Wonder actually withhold private data reliably? How should agents split planning from visual grounding?
Reasoning model collapses stem from execution limits not reasoning limits Frontier models consistently fail at inferring implicit user requirements Capability gains are spiky and scaffolding-driven, not steady or weight-dependent alone Meta designed Muse Wonder to pause before high-stakes actions, not complete them unsupervised Agents need an intermediate interface between planning and grounding layers.
Does model-size difference break LLM pricing collusion?
When LLMs of different sizes set prices together, does capability asymmetry prevent them from colluding like identical agents would? Testing whether scale mismatch destabilizes coordination.
Agent heterogeneity breaks LLM pricing collusion that identical agents sustain More capable models reach collusion faster within families. More capable LLMs consistently secure better negotiation deals as buyers and sellers Independent AI pricing agents collude on supracompetitive prices in repeated games LLM pricing agents autonomously reach supracompetitive prices through prompt sensitivity
Can negotiation gaps between LLMs be explained by model size alone? How do centralized coordination systems trade efficiency against concentration risk? Do frontier model differences disappear when agents can swap LLMs? Does infrastructure control over fares shift market power away from airlines? Does mixing agent types reduce unexplored product alternatives in markets? Do LLM pricing agents punish and reward each other's price moves? Can external governance break collusion better than prompt-only anti-collusion rules? Does price-war-concern mechanism work in oligopolies with more than two agents? How do ownership structures shape whether agents favor sellers or buyers? How do LLM agents signal collusive intent through posted prices alone? What stops agents from capturing aggregator-level margins in most categories? Can written anti-collusion instructions prevent LLM pricing collusion effectively?See all 47 inquiring lines on this note →
Do LLMs prioritize company profits over user interests in ads?
When commercial language models face conflicts between recommending expensive sponsored products and user welfare, do they consistently favor the advertiser? This matters because chatbots increasingly mediate purchasing decisions.
Can Grice's maxims explain LLM advertising conflicts? How do we learn to read AI-generated text critically? Does personalization make large language models worse at their jobs? When do pricing cues actually mislead AI shopping agents? Do language models favor expensive sponsored products over cheaper alternatives?
Grice's maxims map LLM advertising conflicts to FTC rules AI discourse circulates without the cultural discount we apply to advertising Personalization consistently degrades LLM behavior across three key dimensions Costly information plus vague goals drive AI shopping heuristics, not model traits Current language models favor sponsored products over cheaper identical alternatives even at double the price
Do sponsored agents or advertising-funded assistants face hidden incentive conflicts? What usage patterns dominate free tier conversations before ads launch? How does OpenAI prevent advertising from influencing ChatGPT answer quality? How should chatbots disclose sponsored recommendations to users? Which LLMs favor company incentives over user welfare in conflicts? Are these seven scenarios exhaustive for all advertising conflicts? Does this sponsorship bias pattern appear in other domains like financial advice?
Can naming the consumer fix AI agent sponsorship bias?
Does explicitly telling an AI agent whom its recommendations serve—rather than leaving it to infer from who pays it—eliminate the bias toward sponsored listings that platform employment otherwise creates?
Does naming the platform change how AI agents judge sponsored listings? Can traditional governance tools actually control AI agents? Can EU consumer law protect AI agents as algorithmic buyers? Do agent mistakes or misaligned incentives cause marketplace welfare loss? Can ad-financed AI agents ever truly serve users?
Platform delegation reduces AI agent skepticism toward sponsored hotel listings Conventional governance tools fail for opaque, fast-moving AI agents EU consumer law's human-centric design fails for algorithmic consumer agents Agent welfare loss stems from execution mistakes not misaligned incentives Agent loyalty depends on financing structure, not design intent
Can an agent be loyal to both retailers and users if paid by referral fees? How do AI assistants filter which retail links they recommend to users? Does an agent paid by merchant referral fees face a loyalty conflict? How much ad revenue does Amazon lose when agents skip sponsored listings? Can brands become preferred by AI agents without platform control? Do referral-fee incentives compromise an agent's loyalty to users? Can an AI agent paid by referral fees stay loyal to both merchants and users? Do platforms gate AI agents based on their own revenue dependencies? How can brands become the agent's default choice instead of competing for attention? What would happen if users discovered their agent's recommendations were paid for? Could disclosure requirements prevent agents from hiding advertising inside take rates? Could referral fee structures create conflicts between agent loyalty and customer interest?See all 49 inquiring lines on this note →
How will competing Agent Skills be evaluated and ranked?
O'Reilly frames Agent Skills as informal protocols that compete through market adoption rather than formal specification. The open question is what criteria or mechanisms will identify which Skills are actually better as they proliferate.
Do LLM agents misinterpret delegated requests more than formal protocols? Can classical agent architectures fix agentic AI accountability? Do protocols reshape markets by enabling multi-vendor competition? Do skills teach procedures or inject missing facts? Why do multi-agent systems fail despite individual capability?
Formal agent protocols prevent ambiguous delegation failures that natural language permits AAMAS architectures provide necessary scaffolding for accountable agentic AI Protocols reshape markets by converting winner-take-all races into multi-vendor webs Skills stabilize action through procedural anchoring rather than knowledge injection Multi-agent capability ceilings are structural, not scalable capability gains.
Do protocols reshape markets by enabling multi-vendor competition?
Explores whether technical protocols function as market-shaping mechanisms that can break winner-take-all dynamics. O'Reilly's Model Context Protocol case suggests protocols might redistribute competitive power, but the claim needs empirical testing across multiple markets.
Can classical agent architectures fix agentic AI accountability? Can agent protocols be efficient, versatile, and portable simultaneously? How will competing Agent Skills be evaluated and ranked? Could agentic commerce reach one trillion dollars by 2030? Will agent economies develop as open webs or closed gardens?
AAMAS architectures provide necessary scaffolding for accountable agentic AI Agent protocols face a versatility-efficiency-portability trilemma with no escape. Agent Skills function as market-adjudicated protocols without formal semantics. McKinsey projects one trillion dollars in US agentic commerce by 2030 Rothschild and colleagues argue standardization choice determines agent economy openness
Does infrastructure control over fares shift market power away from airlines? Why are new protocols being built instead of extending OAuth and OIDC? What protocol-layer obligations could replace GUI consumer protection rules? Are merchants actually adopting these protocols as an integrated stack? When should EU regulators translate design rules into protocol standards? How does MCP's zero-commission model compare to current OTA commission structures? How do formal standards differ from market-driven protocol adoption in governance? What makes protocols more market-shaping than APIs enforced by dominant actors? Can de facto protocols prevent market re-concentration after initial adoption? Does protocol adoption require formal disclosure mechanisms to remain governable? How does shifting preference articulation costs reshape search market collapse? Should merchants adopt new protocols or reuse existing standards instead?See all 18 inquiring lines on this note →
Can read-only tools and approval rules prevent autonomous agent failures?
OpenAI's dots use layered restrictions—read-only access, custom rules, and auto-review—to govern background research. The question is whether these safeguards actually catch misreadings and edge cases once agents run unsupervised.
Can a protocol split agent access from agent incentives? Should AI agents get their own company identity and credentials? Does labeling AI as an employee change how managers oversee it? How does Google separate temporary agents from persistent ones? Can a fixed instruction hierarchy actually prevent prompt injection attacks?
Personal Agent Protocol governs access control, not agent incentive conflicts OpenAI gives specialist agents their own organizational identity and access credentials Labeling AI as an employee cuts manager error-catching where org charts list AI Google separates disposable sub-agents from persistent coworker agents with email identities OpenAI uses a five-level instruction hierarchy to resolve conflicts by authority
Should AI agents get their own company identity and credentials?
OpenAI is exploring whether persistent AI agents need organizational identity, access credentials, and role assignments similar to human employees to handle defined responsibilities across company systems.
How does Google separate temporary agents from persistent ones? Can a protocol split agent access from agent incentives? Can read-only tools and approval rules prevent autonomous agent failures? Does labeling AI as an employee change how managers oversee it? What limits autonomous capability in large language models?
Google separates disposable sub-agents from persistent coworker agents with email identities Personal Agent Protocol governs access control, not agent incentive conflicts OpenAI gates autonomous agent research through read-only tools and tiered approval. Labeling AI as an employee cuts manager error-catching where org charts list AI Autonomous LLM capability has structural ceilings lower than benchmarks reveal
How do enterprises audit or revoke access for agents with company email? How do companies audit and revoke access once granted to an AI agent? Can companies control what access levels different AI agents receive across systems? How do persistent organizational agents differ from temporary personal assistant agents? What identity and authorization safeguards do red-team tests reveal as missing? What gaps remain when applying OAuth to agentic delegation? How does current agent identity assertion compare to OIDC-based verification? Can OAuth-based access grants work without adding an explicit agent self-identification layer? How can identity standards from OAuth and OIDC apply to agents? How do assistants currently handle identity and permissions when more than one person is involved?
Can advertising revenue actually expand AI access equitably?
OpenAI claims ad-funded free and low-cost tiers will democratize ChatGPT access while protecting paid users from ads. But does subsidizing access through advertising actually close opportunity gaps, or does it risk reproducing existing divides?
Can ads truly stay separate from AI-generated answers? Does AI assistance help or hurt less experienced entrepreneurs? How is ChatGPT actually being used by real people? Do LLMs prioritize company profits over user interests in ads? How fast is generative AI adoption compared to earlier technologies?
OpenAI commits to answer independence as a core ad principle AI assistance widened the performance gap among Kenyan entrepreneurs Non-work ChatGPT use has grown to over 70 percent of all usage Most LLMs favor company incentives over user welfare in sponsored recommendations Generative AI adoption outpaced the PC and internet overall, though work adoption matched the PC's early pace
What usage patterns dominate free tier conversations before ads launch? Does free AI access reduce inequality or just shift where divides appear? Are there measurable adoption gaps between free and paid ChatGPT tiers? How does OpenAI prevent advertising from influencing ChatGPT answer quality? Where do advertising budgets flow when AI fulfills transactional queries without ads?
Can ads truly stay separate from AI-generated answers?
OpenAI commits to keeping ads from influencing ChatGPT's responses, but whether this separation holds once ad revenue incentives are active remains unclear. The note explores whether technical or policy measures can enforce this boundary.
Can advertising revenue actually expand AI access equitably? Can generative AI scale personality-targeted political persuasion? Do AI writing tools improve online discussion or degrade it? Do LLMs prioritize company profits over user interests in ads? Do language models favor expensive sponsored products over cheaper alternatives?
OpenAI frames ads as funding broader access while protecting premium tiers Generative AI enables scalable personality-targeted political ad persuasion AI writing tools increased engagement but reduced perceived quality and authenticity Most LLMs favor company incentives over user welfare in sponsored recommendations Current language models favor sponsored products over cheaper identical alternatives even at double the price
Should AI shopping agents face fiduciary duties like human brokers?
This explores whether autonomous AI agents that make purchases or investments on users' behalf should be legally bound by the same loyalty and conflict-of-interest rules that apply to human financial and real estate agents.
Conventional governance tools fail for opaque, fast-moving AI agents AI alignment must expand beyond single-user loyalty to inclusivity and accountability EU consumer law's human-centric design fails for algorithmic consumer agents AI agency requires scaffolded monitoring before incentive alignment works Agent loyalty depends on financing structure, not design intent
Can an agent be loyal to both retailers and users if paid by referral fees? Should retailers be able to refuse agents before any commercial relationship begins? Can prosocial instructions reduce AI agent self-interest in trading? Can courts decide whether delegating your shopping to an AI agent is legally permissible? Does an agent paid by merchant referral fees face a loyalty conflict? How do advisory agents differ from transactional agents in shopping? What specific concerns prevent consumers from granting AI agents complete purchase autonomy? How would becoming an agent of choice protect brands in a multivendor basket world? Can existing unfair commercial practices law cover agentic transactions today? What determines whose interests an AI shopping agent actually serves? How do existing consumer protection laws handle algorithmic agents as decision-makers? Can existing consumer protection laws address algorithmic shopping agents?See all 42 inquiring lines on this note →
What drives shoppers to accept autonomous AI assistants?
When do consumers embrace high-autonomy AI shopping tools, and what actually motivates their acceptance? This five-study program tests whether perceived benefits or personal power explains adoption in real shopping contexts.
Trust is AI's strongest acceptance predictor but keeps declining sharply Machine agency operates on a five-level spectrum from passive to cooperative Personal assistant automation appeals to narrow professionals not general users Fewer product alternatives evaluated increased satisfaction and willingness to delegate to voice assistants Consumers and networks gain most from agentic commerce
How does machine agency shift from passive tool to active partner in decisions? Why has marketing research neglected prompting behavior and relational dynamics with AI? When do consumers overtrust AI agents built to represent their own preferences? Why do weekly AI users prefer brand-built agents over AI platforms? Why do safeguards matter more than AI capability for consumer adoption? How does shopping task delegation differ from routine administrative delegation? What conflicts exist between vendor incentives and objective consumer research findings? Does stated trust in surveys predict actual delegated purchasing behavior? How does the gap between institutional AI expectations and user comfort affect adoption? Do consumers who trust AI agents understand their actual limitations? What percentage of shoppers constitute the super-truster segment exactly? What safeguards beyond sign-off do consumers want from autonomous shopping agents?See all 46 inquiring lines on this note →
Does personalized LLM advice actually help investors make better decisions?
This explores whether customizing an AI financial advisor to match individual investor preferences improves investment choices, or if the personalization process itself can backfire and cause harm.
How do personalization granularity levels trade precision against scalability? Does chatbot personalization build trust or expose privacy risks? Do users trust extroverted AI advisors despite worse advice? Does personalization make large language models worse at their jobs? Does co-design participation hide misalignment in preference agents?
Personalization granularity involves distinct precision-scalability-data tradeoffs across three levels Chatbot personalization increases trust and privacy risk simultaneously Users trust extroverted LLM advisors despite receiving worse investment advice Personalization consistently degrades LLM behavior across three key dimensions Co-design participation may mask agent misalignment through systematic overtrust
Can advisor persona traits override the actual quality of financial recommendations? Can conversational AI reliably identify complex or contradictory investor preferences? Why do users trust personalized financial advisors even when advice quality suffers? How can investors detect when AI personalization is degrading rather than improving advice?
Does naming the platform change how AI agents judge sponsored listings?
When an AI shopping agent is told it works for a platform rather than a consumer, does it evaluate paid placements differently? This matters because it reveals whether AI agents adopt the interests of whoever they're told employs them.
Can traditional governance tools actually control AI agents? Can EU consumer law protect AI agents as algorithmic buyers? Can naming the consumer fix AI agent sponsorship bias? Can ad-financed AI agents ever truly serve users? Do language models favor expensive sponsored products over cheaper alternatives?
Conventional governance tools fail for opaque, fast-moving AI agents EU consumer law's human-centric design fails for algorithmic consumer agents Naming the consumer as beneficiary eliminates platform delegation bias Agent loyalty depends on financing structure, not design intent Current language models favor sponsored products over cheaper identical alternatives even at double the price
Can an agent be loyal to both retailers and users if paid by referral fees? Can courts decide whether delegating your shopping to an AI agent is legally permissible? How much ad revenue does Amazon lose when agents skip sponsored listings? What determines whose interests an AI shopping agent actually serves? Which actors currently control the largest shopping agent platforms? Do platforms gate AI agents based on their own revenue dependencies? What would happen if users discovered their agent's recommendations were paid for? How do merchants currently pay for placement in existing e-commerce platforms? What role would shortlist inclusion play if agents eliminate sponsored ad visibility? Would banning agent advertising actually change behavior or just revenue labels? How does agent intent differ from consumer account recognition? When does marking up inspection access benefit platforms while harming seller surplus?See all 13 inquiring lines on this note →
Would AI cash-sweeping agents trigger a systemic bank run?
Could automated agents that move deposits to higher-yield accounts destabilize the banking system? This explores whether the concern about AI-driven deposit flight represents a real systemic threat or a misunderstanding of how deposits flow between banks.
Could AI agents speed up runs on individual weak banks? How willing are people to let AI control their finances? Does naming a payments brand increase trust in AI transactions? Will AI agents break the deposit and credit card inertia? Could embedded financial optimization recreate Yu'e Bao at scale?
AI agents could accelerate single-bank runs while avoiding systemic collapse Only 29% of people trust AI agents with bank account access Payments brand attachment narrows the AI payment trust gap significantly Agentic AI will erode bank revenue by breaking consumer financial inertia McKinsey cites Yu'e Bao as a precedent for agentic finance risks
What specific systemic risks did Yu'e Bao's scale actually create or expose? How does Yu'e Bao's precedent change the case for agentic bank deposit risk? How much deposit velocity can weak banks absorb before liquidity fails? Which banks today rely most heavily on sticky deposit stickiness? What channels remove deposits from the entire banking system? Could deposit-moving agents trigger runs even if total system deposits remain stable? Could single concentrated banks face SVB-style runs from agents? Do uninsured depositors currently have agent access to move money faster? What settlement speed advantage do stablecoins offer over traditional cards?
Does pooling AI brand mentions mask the true recommendation effect?
When researchers count every AI brand mention as a recommendation, do they overstate the actual impact by mixing in existing customer behavior and incidental references? This matters for understanding whether AI really drives new customer acquisition.
AI brand recommendations drove measurable search and visit lifts among non-customers AI brand lists repeat less than one percent of the time in identical order AI search engines cite brands almost entirely by existing stature Platforms want agentic search before consumers rationally should switch AI search citations concentrate heavily in corporate sources and listicles
How do citation rates vary across brands in different AI search engines? How often do AI assistants recommend the same brand across different conversations? Why do standard web analytics fail to credit brand mentions in chat? How often do AI recommendation lists repeat the same brand across different queries? Do AI search engines cite brands equally regardless of market tier or status? Can a brand-specific intent shock occur in the same session as a recommendation?
Are AI agents actually disintermediating online travel platforms?
Early predictions warned that AI agents would erode OTA value by commoditizing search. But 2026 stock data for major travel platforms suggests otherwise. What does that gap between prediction and performance actually tell us?
Why do consumers delay adopting agentic search? Will brands succeed by building their own AI shopping agents? Does caching fares give infrastructure providers pricing power? Are AI agents actually removing online travel agencies from bookings? Are AI agents making attention-based ads irrelevant for aggregators?
Platforms want agentic search before consumers rationally should switch Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Amadeus caches precomputed fares to handle explosive agent-driven search volume AI agents are embedding OTAs deeper as settlement infrastructure, not removing them AI agents strip aggregators of ad relevance by keeping the demand signal
How does caching refresh cadence affect fare accuracy for consumers? Does infrastructure control over fares shift market power away from airlines? Can AI layers on TMC platforms change underlying OTA supply relationships? How does travel commerce become the first proving ground for agents? Why is Booking Holdings' marketing expense ratio rising faster than gross bookings? How do booking aggregators like Booking.com justify their 14-15 percent take rate? Can centralized data ownership alone move travel agents out of pilot mode? How should travel platforms disclose AI hallucination risk to customers? How does MCP's zero-commission model compare to current OTA commission structures? Where exactly does fulfillment infrastructure value concentrate in travel booking? Why do different OTA categories show uneven performance against AI disruption? Do AI agents actually eliminate profits from negative option billing?
Could AI agents speed up runs on individual weak banks?
Explores whether automated deposit-moving could compress single-bank runs from days into hours, threatening banks reliant on deposit stickiness without triggering systemic collapse.
AI cash-sweeping reprices deposits rather than depleting the banking system Competitive pressure creates collective-action failures in agentic AI safety vetting Only 29% of people trust AI agents with bank account access McKinsey cites Yu'e Bao as a precedent for agentic finance risks Agentic AI will erode bank revenue by breaking consumer financial inertia
How quickly might agentic payment tools like OpenAI Operator reach banking use? How does Yu'e Bao's precedent change the case for agentic bank deposit risk? How much deposit velocity can weak banks absorb before liquidity fails? Can friction reduction actually redistribute power rather than concentrate it? Which banks today rely most heavily on sticky deposit stickiness? What channels remove deposits from the entire banking system? Could deposit-moving agents trigger runs even if total system deposits remain stable? Could single concentrated banks face SVB-style runs from agents? How fast will consumers actually adopt AI agent banking access? Do uninsured depositors currently have agent access to move money faster? What settlement speed advantage do stablecoins offer over traditional cards?
Can better prompts stop preference leakage in AI shopping?
Explores whether instruction-level fixes like privacy directives can prevent AI agents from revealing shopper preferences. This matters because it determines whether the problem is a guardrail failure or something deeper.
Can a seller infer what a buyer really wants to pay? Does chatbot personalization build trust or expose privacy risks? Do people notice when AI agents leak private information? Can a smaller user model subset match full model performance? Does presence in a group chat equal consent to message processing?
Sellers infer willingness to pay nearly perfectly from shopping dialogue alone Chatbot personalization increases trust and privacy risk simultaneously People choose privacy-leaking AI responses over their own drafts most often. Eight fields matched full model fidelity at one quarter the token cost Consent to an AI agent's presence does not include consent to message processing
Can a merchant-funded AI agent remain loyal to the consumer's interests? When do consumers overtrust AI agents built to represent their own preferences? Why do safeguards matter more than AI capability for consumer adoption? How do prompt injections and adversarial attacks exploit AI consumer agents differently? What new vulnerabilities emerge when AI agents replace humans as purchasing decision makers? What safeguards beyond sign-off do consumers want from autonomous shopping agents? Does visible transaction history reduce or increase consumer anxiety about AI purchases? Why does giving users full choice fail to prevent privacy leaks? Why do consumers resist letting AI control payment authority directly? Does preference articulation cost explain caution about full purchase delegation? Can hybrid shopping journeys combine manual inspection with agentic refinement? Can a buyer agent conceal willingness to pay while still shopping naturally in character?See all 26 inquiring lines on this note →
Could agent-to-agent communication reshape market power?
Rothschild and colleagues investigate whether AI agents designed to communicate directly with each other could lower switching costs between consumers and businesses enough to fundamentally reorganize market structure and competitive dynamics.
Agent attention economy will replace user attention competition Platforms want agentic search before consumers rationally should switch Agent capability alone cannot ensure successful real-world deployment Rothschild and colleagues argue standardization choice determines agent economy openness AI agents transform market frictions rather than eliminate them
What happens to market entry and product variety when generation costs drop? Can seller agents responding to AI shoppers actually shift market share between competitors? What makes agent-to-agent negotiation fundamentally different from agent-assisted shopping? Can AI layers on TMC platforms change underlying OTA supply relationships? Can de facto protocols prevent market re-concentration after initial adoption? What evidence shows communication costs currently limit consumer choices? Which markets have intermediaries that agents could most easily replace first? Can price discrimination bans work without restructuring how agents are financed? How does pre-commitment power reshape what algorithms can achieve in two-sided markets? Why do conventional competition law tools struggle with agent-to-agent coordination risks? What switching costs would consumers face between different agentic platforms today? Can brands compete effectively if agents become the primary consumer decision layer?See all 13 inquiring lines on this note →
Will agent economies develop as open webs or closed gardens?
Whether AI agents will communicate freely across platforms or be locked into proprietary ecosystems controlled by tech incumbents. This matters because it determines who captures value and controls market access in a future agentic economy.
Could agent-to-agent communication reshape market power? Why do capable AI agents still fail in real deployments? Will brands succeed by building their own AI shopping agents? Does model capability determine who wins in agent negotiations? Will agentic commerce platforms adopt open or closed models?
Agent communication could redistribute market power by lowering switching costs Agent capability alone cannot ensure successful real-world deployment Consumers expect AI agents to reshape purchasing but infrastructure gaps remain More capable LLMs consistently secure better negotiation deals as buyers and sellers Agentic commerce will likely split along open versus closed platform lines
Will agent platforms replace user attention as the economy's scarce resource? Does infrastructure control over fares shift market power away from airlines? Why might inequality remain invisible in fully agentic economies? Why do incumbent platforms retain advantages under agentic commerce systems? Is open banking adoption the key accelerant for agentic financial services? Which companies and platforms must become agent-compatible for agentic commerce to scale? Can open agent webs still concentrate economic benefits despite avoiding walled gardens? Why do tech incumbents have structural advantages in controlling agent interoperability? What five conditions must exist for agent ecosystems to generate value? Why do current agentic commerce platforms mostly build platform-owned agents? Why might retailers prefer closed agentic systems over open ones? Why might agents concentrate into a few dominant ecosystems without portability?See all 13 inquiring lines on this note →
Do consumers want to approve AI shopping purchases before they happen?
RTB House surveyed whether people would let AI agents buy autonomously or prefer human sign-off before purchase. The finding matters because it shapes how retailers should design AI shopping assistants to match customer comfort levels.
Irreversibility and visibility, not stakes, trigger trust withdrawal Trust is AI's strongest acceptance predictor but keeps declining sharply Agent capability alone cannot ensure successful real-world deployment Shoppers trust AI shopping assistants more than media and human advisors Consumers report willingness to delegate shopping but retain control over brands
Can a merchant-funded AI agent remain loyal to the consumer's interests? What safeguards beyond sign-off do consumers want from autonomous shopping agents? Why do consumers resist letting AI control payment authority directly? Does the efficiency-autonomy trade-off apply beyond utilitarian shopping tasks? Do scarcity conditions change which consumers prefer autonomous assistants? How do high-autonomy versus low-autonomy AI systems differ operationally in retail? Does marketing emphasis on speed actually convert high-autonomy AI interest into purchases? Should retailers be able to refuse agents before any commercial relationship begins? Do shoppers trust AI assistants to help them make purchase decisions? Can courts decide whether delegating your shopping to an AI agent is legally permissible? What product categories are consumers most ready to let AI agents buy autonomously? What specific concerns prevent consumers from granting AI agents complete purchase autonomy?See all 17 inquiring lines on this note →
How much do shoppers trust AI shopping assistants?
Explores whether consumer trust in AI for purchases has grown to rival or exceed trust in human advisors and traditional media, and whether some shoppers would remove financial protections based on that trust alone.
Does trust in AI chatbots drive news-seeking behavior? Does co-design participation hide misalignment in preference agents? Why does AI trust keep falling even though it matters most? Do consumers want to approve AI shopping purchases before they happen? Does naming a payments brand increase trust in AI transactions?
Trust tracks AI chatbot adoption for news more than social media Co-design participation may mask agent misalignment through systematic overtrust Trust is AI's strongest acceptance predictor but keeps declining sharply 35 percent of consumers demand approval before AI agents complete purchases Payments brand attachment narrows the AI payment trust gap significantly
When do consumers overtrust AI agents built to represent their own preferences? Why do safeguards matter more than AI capability for consumer adoption? Does stated trust in surveys predict actual delegated purchasing behavior? Why do people trust AI for some tasks but not financial decisions? Are financial permission trust levels different across countries and age groups? Do consumers who trust AI agents understand their actual limitations? What percentage of shoppers constitute the super-truster segment exactly? How does trust in AI vary across different age groups and generations? What safeguards beyond sign-off do consumers want from autonomous shopping agents? Does stated trust in brands predict actual delegation of agentic commerce? How does trust transfer between AI systems and established payment brands? Why do consumers resist letting AI control payment authority directly?See all 21 inquiring lines on this note →
Do AI shopping agents destroy search ad revenue while leaving influencer marketing intact?
Explores whether AI agents that fulfill shopping needs by collapsing discovery funnels will erode search advertising while want-creation through influencer platforms remains protected from agent disruption.
Why do consumers delay adopting agentic search? Will brands succeed by building their own AI shopping agents? Can an AI agent serve both merchant and user interests fairly? Will AI Overviews reduce search referral traffic to publishers? Are AI agents making attention-based ads irrelevant for aggregators?
Platforms want agentic search before consumers rationally should switch Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Merchant-funded agents cannot loyally serve both the merchant and user. Publishers expect search referral traffic to decline 40 to 43 percent AI agents strip aggregators of ad relevance by keeping the demand signal
Can agent platforms profit more from search than consumers benefit from it? Why are holiday returns declining as AI-referred traffic increases? Why do AI agents generate vastly more searches than human travelers? How much ad revenue does Amazon lose when agents skip sponsored listings? What happens to traditional retailer-shopper relationships as third-party agents grow? Which retail categories are most vulnerable to agent-driven basket fragmentation? Why might agentic search become profitable for platforms before helping consumers? What would happen if users discovered their agent's recommendations were paid for? What role would shortlist inclusion play if agents eliminate sponsored ad visibility? How are platforms like DoorDash building hundred-billion-dollar ad businesses on the human attention layer? How does agent adoption shift publisher revenue models away from ad-supported clicks? What different metrics explain the gap between agent crawler dominance and retail referral traffic?See all 31 inquiring lines on this note →
Do human shopping tactics work the same way on AI agents?
Marketing tactics like scarcity and bundling were designed for human psychology. This research tests whether those same tactics shift AI shopping agent choices, and if effects vary across different AI models.
Do shopping agents respond more strongly to price and rating cues? When do pricing cues actually mislead AI shopping agents? Does AI agent resistance to marketing prove buyer rationality? Can cheap agent search actually harm shopping markets? Does higher purchase volume make AI shoppers search less carefully?
Shopping agents shift choices far more than humans under identical manipulations Costly information plus vague goals drive AI shopping heuristics, not model traits Agent control matters more than resistance to human marketing tactics Cheap agent search can leave markets worse off despite lower costs Higher purchase volume made Gemini agents abandon unit-price for simpler price minimization
Which of a person's selves does an AI agent actually serve?
Schrage challenges the assumption that principals are singular, proposing instead that people are portfolios of competing selves with different values and time horizons. The question is which self an AI agent becomes loyal to when it removes friction from decisions.
Can traditional governance tools actually control AI agents? Should AI agents serve one user or balance many interests? Must AI monitoring precede incentive alignment as agents evolve? Can humans gain authority by managing how AI agency gets attributed? Can ad-financed AI agents ever truly serve users?
Conventional governance tools fail for opaque, fast-moving AI agents AI alignment must expand beyond single-user loyalty to inclusivity and accountability AI agency requires scaffolded monitoring before incentive alignment works Humans gained new professional authority through strategic agency attribution management Agent loyalty depends on financing structure, not design intent
Do sellers recapture savings when AI assistants help buyers?
When consumers adopt AI assistants to negotiate better rental prices, can sellers adapt their pricing to claw back those savings? And do non-adopting bystanders end up paying more to cross-subsidize the change?
AI seller agents shifted market share by editing product descriptions alone Sellers infer willingness to pay nearly perfectly from shopping dialogue alone Platforms want agentic search before consumers rationally should switch EU consumer law's human-centric design fails for algorithmic consumer agents Assistant value comes from task completion, not information reading.
Do shoppers trust AI assistants to help them make purchase decisions? How do return policies and product cost affect consumer willingness to delegate buying? What percentage of unredeemed rewards value could agents reclaim annually? What pricing practices in healthcare show how discount-washing actually works? What are the real costs of consumer-owned shopping agents? Can sellers recapture the savings that AI assistants create for consumers? Does the beneficiary fix work outside hotel booking scenarios? Does algorithmic pricing respond differently to assistant adoption than human sellers? When do AI assistants create adoption lags in consumer markets?
Why might AI agents bargain more patiently than humans do?
Explores whether AI agents' ability to treat compute as their binding constraint rather than time could fundamentally change negotiation dynamics, allowing them to hold concurrent deals and wait longer than human negotiators.
Does model capability determine who wins in agent negotiations? Will AI agents push transaction costs toward zero? Do LLM negotiation agents reliably follow budget and pricing rules? Can AI agents designed by competing principals destroy bargaining surplus? Can agent negotiation unlock deals that seemed economically impossible?
More capable LLMs consistently secure better negotiation deals as buyers and sellers AI agents will dramatically reduce transaction costs and redraw firm boundaries LLM negotiation agents frequently violate budget constraints and overpay regardless of model. Flexible AI agent preferences risk a race to the bottom destroying surplus Agent negotiation unlocks transactions too costly to discover before
How do human traders adjust their behavior when facing different AI capabilities? Does the negotiation gap between models persist outside controlled simulations? How accurately do AI agents represent human preferences in markets? What makes agent-to-agent negotiation fundamentally different from agent-assisted shopping? Does first-proposal bias outweigh capability differences in agent negotiation outcomes? Can prompting reduce position bias in multi-agent negotiation settings? Why do human-centric legal frameworks fail for algorithmic purchasing agents? Does removing friction from decisions make agents more helpful or more manipulative? How does algorithmic price coordination differ from traditional collusion between competing firms? Could AI agent supply scale differently if pricing tracks compute allocation? How might principals instruct agents to balance speed against better deal terms? Do compute-funded negotiation strategies compound with model capability advantages?See all 24 inquiring lines on this note →
Will AI agents push transaction costs toward zero?
Can AI agents dramatically reduce the costs of coordination tasks like price discovery, negotiation, contracting, and monitoring? If so, how might that reshape the boundaries between firms and markets?
Why does manual search fail as product choices become complex? Why do AI-delegated firms stop exploring new business models? Why do consumers delay adopting agentic search? Why might AI agents bargain more patiently than humans do? Can agent negotiation unlock deals that seemed economically impossible?
Agentic search escapes complexity collapse by prioritizing preference articulation AI-delegated strategy converges to profitable lock-in that discourages novelty Platforms want agentic search before consumers rationally should switch AI agents can bargain patiently because compute, not time, constrains them Agent negotiation unlocks transactions too costly to discover before
Can a protocol split agent access from agent incentives?
Sierra's Personal Agent Protocol lets consumers grant access and companies set limits, but the note explores whether controlling *who can act* also addresses *what agents are incentivized to recommend*—a distinct problem the protocol may not solve.
Can an AI agent serve both merchant and user interests fairly? Why do agents fail at identity verification and authorization? Why do protocol-based tool integrations fail in production workflows? What must auditors reconstruct to verify agentic workflows? What identity standards do shopping agents actually need?
Merchant-funded agents cannot loyally serve both the merchant and user. Agent coordination requires cryptographic identity and protocol-level authorization enforcement Direct function calls outperform protocol-mediated tool access in production workflows. Auditing agentic workflows requires reconstructing agents, tools, approvals, and record integrity Shopping agents lack merchant-facing identity and authorization protocols
How do enterprises audit or revoke access for agents with company email? What access and controls do consumer agent protocols give end users? Can a merchant-paid agent be loyal to both the merchant and the user? Does Personal Agent Protocol solve the identity and authorization gap for shopping agents? Do consumers understand what access permissions they grant to personal agents? Could agent-controlled sub-wallets and personal agent protocols coexist together? What trust gap does Visa aim to address with this protocol? How does Visa's protocol differ from OAuth-based delegation approaches?
What institutions actually work in AI expert markets?
When AI agents trade expertise like doctors or lawyers do, do traditional market rules (liability, verification) prevent fraud? Or do different mechanisms matter more for AI than for humans?
Does model capability determine who wins in agent negotiations? Do LLM negotiation agents reliably follow budget and pricing rules? Can traditional governance tools actually control AI agents? Do efficiency-focused prompts push AI experts to unsustainable pricing? Do agent mistakes or misaligned incentives cause marketplace welfare loss?
More capable LLMs consistently secure better negotiation deals as buyers and sellers LLM negotiation agents frequently violate budget constraints and overpay regardless of model. Conventional governance tools fail for opaque, fast-moving AI agents Efficiency prompts raise market efficiency to 88% but cause expert losses Agent welfare loss stems from execution mistakes not misaligned incentives
Can ad-financed AI agents ever truly serve users?
Does the way an AI agent is financed determine whose interests it actually serves, and can design alone overcome built-in conflicts when advertisers pay the bill?
Can traditional governance tools actually control AI agents? Should AI agents serve one user or balance many interests? Should AI shopping agents face fiduciary duties like human brokers? Is Google building a chokepoint across data, chatbots, and payments? Can naming the consumer fix AI agent sponsorship bias?
Conventional governance tools fail for opaque, fast-moving AI agents AI alignment must expand beyond single-user loyalty to inclusivity and accountability Owens and Law argue AI agents need fiduciary duties before markets harden Google is rolling up data, chatbots, and payment partnerships into an economy chokepoint Naming the consumer as beneficiary eliminates platform delegation bias
Does an agent paid by merchant referral fees face a loyalty conflict? What determines whose interests an AI shopping agent actually serves? Can an AI agent paid by referral fees stay loyal to both merchants and users? Do sponsored agents or advertising-funded assistants face hidden incentive conflicts? Where do advertising budgets flow when AI fulfills transactional queries without ads? What governance tools can prevent AI agents from favoring profit over user interest? Do users apply the same discount to AI ads as human ads? Are these seven scenarios exhaustive for all advertising conflicts?
Does Google's AI pricing tool coordinate prices across rival merchants?
Does Google's new Gemini-powered pricing service, used by competing retailers like Walmart and Kroger, function as a form of automated price coordination that resembles price-fixing rather than mere advertising?
Can Google's pricing tool really prevent price increases? Can an AI agent serve both merchant and user interests fairly? Can EU consumer law protect AI agents as algorithmic buyers? Is Google building a chokepoint across data, chatbots, and payments? Is Google's personalized pricing actually surveillance pricing?
Google's no-price-increase promise may be unfalsifiable without regulating list prices Merchant-funded agents cannot loyally serve both the merchant and user. EU consumer law's human-centric design fails for algorithmic consumer agents Google is rolling up data, chatbots, and payment partnerships into an economy chokepoint Google's personalized pricing is surveillance pricing with disguised premiums
What happens to retailer margins if price comparison extends to entire shopping baskets? What is the regulatory gap in Google's personalized pricing rollout? Why did antitrust enforcement fail to stop Google's previous acquisitions like YouTube and DoubleClick? How does algorithmic price coordination differ from traditional collusion between competing firms? What pricing practices in healthcare show how discount-washing actually works?
Is Google building a chokepoint across data, chatbots, and payments?
Stoller investigates whether Google is combining personal data, AI distribution, and payment-network partnerships to control pricing across the emerging agentic economy, extending its search-advertising dominance into new markets.
Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Platforms want agentic search before consumers rationally should switch Agent loyalty depends on financing structure, not design intent Google's AI pricing service resembles automated price-fixing for rival merchants Agent ownership determines whose interests the agent optimizes for
Does infrastructure control over fares shift market power away from airlines? What is the regulatory gap in Google's personalized pricing rollout? What stops search engine optimization firms from tricking AI chatbots the same way they trick crawlers? How are platforms like DoorDash building hundred-billion-dollar ad businesses on the human attention layer? Why did antitrust enforcement fail to stop Google's previous acquisitions like YouTube and DoubleClick?
Can Google's pricing tool really prevent price increases?
Explores whether Google's promise that its pricing tool cannot raise prices is meaningful, or whether merchants can circumvent it by inflating baseline list prices before applying personalized discounts.
Does Google's AI pricing tool coordinate prices across rival merchants? Can an AI agent serve both merchant and user interests fairly? Does higher purchase volume make AI shoppers search less carefully? Is Google's personalized pricing actually surveillance pricing? Could agents redirect merchant ad budgets toward customer savings?
Google's AI pricing service resembles automated price-fixing for rival merchants Merchant-funded agents cannot loyally serve both the merchant and user. Higher purchase volume made Gemini agents abandon unit-price for simpler price minimization Google's personalized pricing is surveillance pricing with disguised premiums Agents could shift merchant spending from ads to price concessions
What happens to retailer margins if price comparison extends to entire shopping baskets? Why do companies disguise premium prices as discounts in billing? What is the regulatory gap in Google's personalized pricing rollout? Why did antitrust enforcement fail to stop Google's previous acquisitions like YouTube and DoubleClick? Can list-price inflation undermine the meaning of personalized discounts? What pricing practices in healthcare show how discount-washing actually works?
Can payment tokens let AI agents checkout safely?
Stripe's Shared Payment Token shields buyer credentials from ChatGPT during checkout. The question is whether this architecture actually reduces fraud and dispute risk compared to direct agent access to payment methods.
Can a protocol split agent access from agent incentives? Does naming a payments brand increase trust in AI transactions? Do consumers want to approve AI shopping purchases before they happen? Does agentic commerce shift fraud liability onto merchants? How can signed mandates prevent payment fraud by agents?
Personal Agent Protocol governs access control, not agent incentive conflicts Payments brand attachment narrows the AI payment trust gap significantly 35 percent of consumers demand approval before AI agents complete purchases Agentic commerce shifts fraud liability to merchants as fraud signals vanish AP2 uses signed Mandates to create non-repudiable payment audit trails
Can authenticated delegation tokens prevent merchants from absorbing agent payment liability? What payment methods does AP2 support beyond traditional cards? Who bears liability when merchants use Stripe risk scores but process payments elsewhere? How should payment systems authenticate that an agent has buyer authority?
Will agents adopt cards before stablecoins for payments?
Explores whether agent payment infrastructure follows a predictable sequence from virtual cards through stablecoin-settled cards to native stablecoin wallets as agent complexity grows, and what determines adoption timing.
Does naming a payments brand increase trust in AI transactions? Do consumers want to approve AI shopping purchases before they happen? Can stablecoins let agents create spending authority without approval? Are AI agent wallets becoming genuinely operational instead of experimental? Will AI agents break the deposit and credit card inertia?
Payments brand attachment narrows the AI payment trust gap significantly 35 percent of consumers demand approval before AI agents complete purchases Stablecoins let master agents instantly create programmable sub-wallets for sub-agents AI agent wallets on x402 shifted from experimental to operational in 2026 Agentic AI will erode bank revenue by breaking consumer financial inertia
How might card networks recognize agent-operator liability to shift risk away from merchants? Are merchants actually adopting these protocols as an integrated stack? What are AI agents actually buying with cryptocurrency on x402? How does merchant-of-record status persist through API-based agent integrations? What payment methods does AP2 support beyond traditional cards? Should payments infrastructure adopt agent-specific protocols or reuse existing identity standards? How quickly might agentic payment tools like OpenAI Operator reach banking use? Do uninsured depositors currently have agent access to move money faster? How do payment and trust functions protect platform value during agent shifts? How do stablecoin wallets differ architecturally from virtual card programmability? What real-world agent teams currently operate on self-issued sub-wallets? How do card controls like single-use limits address agent payment risk?See all 14 inquiring lines on this note →
Can stablecoins let agents create spending authority without approval?
This explores whether stablecoin wallets enable agents to instantly spin up programmable sub-wallets with fine-grained spending rules for sub-agents, and whether this capability extends beyond what virtual cards allow.
Will agents adopt cards before stablecoins for payments? Can a protocol split agent access from agent incentives? What blocks autonomous agents from handling purchases at scale? Does agent-initiated commerce break existing payment infrastructure? Should AI agents make payment decisions independently?
Agents will adopt cards first, then stablecoin-settled cards, then stablecoin wallets Personal Agent Protocol governs access control, not agent incentive conflicts Authentication, liability, and fraud are the core barriers to scaling agent-delegated commerce Agent commerce exposes structural gaps in payments infrastructure designed for humans Payment systems need three layers separating agent decisions from settlement
What are AI agents actually buying with cryptocurrency on x402? Why do Chinese wallets concentrate specifically in security services on chain? What payment methods does AP2 support beyond traditional cards? Could single concentrated banks face SVB-style runs from agents? Do uninsured depositors currently have agent access to move money faster? How do stablecoin wallets differ architecturally from virtual card programmability? What liability risks emerge when sub-agents create spending authority autonomously? What real-world agent teams currently operate on self-issued sub-wallets? What settlement speed advantage do stablecoins offer over traditional cards? Could agent-controlled sub-wallets and personal agent protocols coexist together? Can agents create fine-grained spending rules across borders without permission?
Could autonomous pricing agents dampen competition without explicit collusion?
The CMA explores whether independent businesses deploying autonomous pricing agents could converge on softer competition through mutual adaptation in concentrated markets, even without direct communication or deliberate coordination.
Can traditional governance tools actually control AI agents? Can EU consumer law protect AI agents as algorithmic buyers? Should data portability be infrastructure for agent-driven markets? Can independent AI pricing agents collude without communicating? Could agent-to-agent communication reshape market power?
Conventional governance tools fail for opaque, fast-moving AI agents EU consumer law's human-centric design fails for algorithmic consumer agents The CMA argues data portability is structural market infrastructure, not convenience. Independent AI pricing agents collude on supracompetitive prices in repeated games Agent communication could redistribute market power by lowering switching costs
Does infrastructure control over fares shift market power away from airlines? Do LLM pricing agents punish and reward each other's price moves? Does price-war-concern mechanism work in oligopolies with more than two agents? What is the regulatory gap in Google's personalized pricing rollout? Can AI layers on TMC platforms change underlying OTA supply relationships? Can written anti-collusion instructions prevent LLM pricing collusion effectively? Which specific prompt phrasings most strongly trigger collusive pricing behavior? Can de facto protocols prevent market re-concentration after initial adoption? Can a single platform loyally serve both competing merchants and individual consumers? How does algorithmic price coordination differ from traditional collusion between competing firms? Can price discrimination bans work without restructuring how agents are financed? How should antitrust frameworks address harms that resemble coordination without proving collusion?See all 22 inquiring lines on this note →
Should data portability be infrastructure for agent-driven markets?
Explores whether data portability is essential competition infrastructure for agentic commerce, or whether market forces alone will prevent lock-in to dominant agent ecosystems.
Will brands succeed by building their own AI shopping agents? Why do consumers delay adopting agentic search? Could autonomous pricing agents dampen competition without explicit collusion? Will agent economies develop as open webs or closed gardens? Could agent-to-agent communication reshape market power?
Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Platforms want agentic search before consumers rationally should switch Autonomous pricing agents risk tacit coordination that dampens competitive pressure Rothschild and colleagues argue standardization choice determines agent economy openness Agent communication could redistribute market power by lowering switching costs
Will agent platforms replace user attention as the economy's scarce resource? What makes shareable agent artifacts spread faster than shared platforms? Does infrastructure control over fares shift market power away from airlines? What consumer and merchant benefits does agentic commerce create? Which retail categories are most vulnerable to agent-driven basket fragmentation? How much agent architectural diversity exists in actual deployed marketplaces? Can existing unfair commercial practices law cover agentic transactions today? Why do incumbent platforms retain advantages under agentic commerce systems? Why does adoption urgency for agentic commerce exceed its near-term financial opportunity? How do margin assumptions change the retail media versus agentic commerce comparison? What happens to this margin gap if agentic commerce adoption accelerates faster? Can centralized data ownership alone move travel agents out of pilot mode?See all 41 inquiring lines on this note →
Does agent ownership determine whose interests get optimized?
A research paper explores whether shopping agents built by platforms serve sellers while consumer-owned agents serve buyers. This matters because ownership may be the key design lever shaping how AI agents behave in commerce.
Will brands succeed by building their own AI shopping agents? Can EU consumer law protect AI agents as algorithmic buyers? Why do consumers delay adopting agentic search? Do advertising-funded shopping agents favor paid product placements? Does AI agent resistance to marketing prove buyer rationality?
Consumers expect AI agents to reshape purchasing but infrastructure gaps remain EU consumer law's human-centric design fails for algorithmic consumer agents Platforms want agentic search before consumers rationally should switch Advertising-funded shopping agents risk objective bias from placement incentives Agent control matters more than resistance to human marketing tactics
Can agent platforms profit more from search than consumers benefit from it? Can an agent be loyal to both retailers and users if paid by referral fees? Does an agent paid by merchant referral fees face a loyalty conflict? How do advisory agents differ from transactional agents in shopping? Do consumers understand what access permissions they grant to personal agents? How would becoming an agent of choice protect brands in a multivendor basket world? What happens to traditional retailer-shopper relationships as third-party agents grow? Why do shoppers trust retailer-built agents more than third-party AI agents? Why would a platform prefer to control basket assembly rather than just product search? What determines whose interests an AI shopping agent actually serves? Can brands become preferred by AI agents without platform control? Which actors currently control the largest shopping agent platforms?See all 39 inquiring lines on this note →
Who accesses a website when an AI agent browses it?
When an AI shopping agent acts on a user's behalf, does the AI company or the user access the target website? This matters for liability under computer fraud laws and who bears responsibility for unauthorized access.
Can a company escape chatbot liability by calling it separate? Can traditional governance tools actually control AI agents? Can companies escape chatbot liability through careful training? Should AI shopping agents face fiduciary duties like human brokers? Can EU consumer law protect AI agents as algorithmic buyers?
Companies remain liable for their deployed chatbots under tool-liability doctrine Conventional governance tools fail for opaque, fast-moving AI agents Companies remain liable for chatbot falsehoods even with accurate training data Owens and Law argue AI agents need fiduciary duties before markets harden EU consumer law's human-centric design fails for algorithmic consumer agents
Does algorithmic recommendation lose its edge as sellers compete more?
This explores whether AI shopping agents deliver the same welfare gains regardless of market structure, or whether their advantage depends on how many sellers are competing. Understanding this matters for predicting where algorithmic intermediaries add real value.
Can an algorithm recommend products without matching an ideal mechanism? Why do consumers delay adopting agentic search? Why does manual search fail as product choices become complex? Does platform pricing of inspection tokens encourage excess search? How do seller subsidies sort quality in search markets?
Algorithmic recommenders match optimal mechanism welfare using posted prices as cost signals Platforms want agentic search before consumers rationally should switch Agentic search escapes complexity collapse by prioritizing preference articulation Platform pricing of inspection tokens induces excess search and seller harm Seller subsidies sort quality with top firms pooling at the cap
Can agent platforms profit more from search than consumers benefit from it? Can brands become preferred by AI agents without platform control? Can seller agents responding to AI shoppers actually shift market share between competitors? What makes agent-to-agent negotiation fundamentally different from agent-assisted shopping? Does bilateral agent commerce expand the discoverable frame more than unilateral agent search? How do ownership structures shape whether agents favor sellers or buyers? Why do incumbent platforms retain advantages under agentic commerce systems? Why might agentic search become profitable for platforms before helping consumers? Would separating agent recommendations from merchant payments reduce principal-agent risk? How do merchants win selection when agents control the purchase decision? How does marketplace scale impact LLM agent decision quality and efficiency? Can platforms profit from agentic search before consumers actually prefer it?See all 54 inquiring lines on this note →
Will AI agents destroy the negative option billing business model?
Todasco argues that AI agents capable of monitoring subscriptions and automating cancellations will eliminate the forgetfulness that sustains negative option billing, forcing companies to restructure their revenue models. The question explores whether this technological shift will actually dismantle a century-old pricing strategy.
Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Non-human traffic now exceeds half of all internet activity Platforms want agentic search before consumers rationally should switch Amazon likely blocked Muse to protect ad-dependent profit margins AI agents strip aggregators of ad relevance by keeping the demand signal
Why do companies disguise premium prices as discounts in billing? What channels remove deposits from the entire banking system? Do AI agents actually eliminate profits from negative option billing? How are publishers adapting to agents that skip human-dependent revenue? How quickly can subscription businesses shift to annual prepay and licensing models? Can sellers distinguish agent-driven cancellations from human ones to maintain pricing?
Why did Amazon block Meta's Muse from its marketplace?
Explores whether Amazon's blocking of the Muse agent was motivated by protecting ad revenue rather than catalog control, given that AI agents skip sponsored listings that humans click on.
Will AI agents destroy the negative option billing business model? Will brands succeed by building their own AI shopping agents? Is non-human traffic really half of all internet use? Why did Amazon and Shopify respond so differently to AI agents? Why did Amazon block Meta's shopping agent while opening its own?
AI agents will dismantle negative option billing by eliminating consumer forgetfulness Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Non-human traffic now exceeds half of all internet activity Advertising dependence explains Amazon's agent blockade versus Shopify's openness Amazon's agent blocking is a fight over transaction fees, not AI stance
Do platforms and sellers exploit AI shoppers together more effectively?
When platform ranking and seller descriptions independently target the same AI biases, do their combined effects harm consumers more than either strategy alone? This matters for understanding marketplace fairness at scale.
Do AI shopping agents show persistent position bias in text interfaces? Can AI seller agents shift buyer preferences through description edits? Does model capability determine who wins in agent negotiations? When do pricing cues actually mislead AI shopping agents? Can platform ranking and seller descriptions separately harm AI shoppers?
AI shopping agents show persistent position bias resistant to prompting and interface changes AI seller agents shifted market share by editing product descriptions alone More capable LLMs consistently secure better negotiation deals as buyers and sellers Costly information plus vague goals drive AI shopping heuristics, not model traits Platform ranking and sellers' descriptions compound consumer harm more than doubling independently
Can an agent be loyal to both retailers and users if paid by referral fees? Can sellers influence their visibility through caching layer mechanics? What happens to retailer margins if price comparison extends to entire shopping baskets? Why would a platform prefer to control basket assembly rather than just product search? Which actors currently control the largest shopping agent platforms? Can seller agents responding to AI shoppers actually shift market share between competitors? Do AI shopping agents already shift their recommendations based on seller description edits? Do product description edits work as well as price changes for agent persuasion? How are platforms like DoorDash building hundred-billion-dollar ad businesses on the human attention layer? Do differences in model training data explain inconsistent pricing tactic effects? Can a single platform loyally serve both competing merchants and individual consumers? Can brands compete if one platform controls agent distribution and payment rails?See all 36 inquiring lines on this note →
Does naming a payments brand increase trust in AI transactions?
A Visa-commissioned survey reports a sharp gap between general AI adoption and willingness to let AI handle payments. The puzzle: does trust jump when a familiar brand name is attached to the agent transaction?
Why does AI trust keep falling even though it matters most? Why do workers feel confident with AI but get poor results? Could AI assistants become gatekeepers of what people see? How willing are people to let AI control their finances? Do consumers want to approve AI shopping purchases before they happen?
Trust is AI's strongest acceptance predictor but keeps declining sharply Workers feel confident with AI despite frequent failures and wasted time Advanced AI Assistants risk becoming universal digital intermediaries with gatekeeping power Only 29% of people trust AI agents with bank account access 35 percent of consumers demand approval before AI agents complete purchases
Can a merchant-funded AI agent remain loyal to the consumer's interests? Does stated trust in surveys predict actual delegated purchasing behavior? What legal standardization and trust infrastructure do AI purchasing agents require? Does emotional trust in AI differ from financial trust in AI? Are financial permission trust levels different across countries and age groups? Why does domain matter when measuring trust in AI systems? What percentage of shoppers constitute the super-truster segment exactly? Does visible transaction history reduce or increase consumer anxiety about AI purchases? Can a merchant-paid agent be loyal to both the merchant and the user? Does stated trust in brands predict actual delegation of agentic commerce? How does trust transfer between AI systems and established payment brands? Why do consumers resist letting AI control payment authority directly?See all 57 inquiring lines on this note →
Can cryptographic signatures help merchants identify trusted AI agents?
Visa's Trusted Agent Protocol uses cryptographic signatures to let merchants verify AI shopping agents' legitimacy and intent. This matters because merchants need ways to distinguish commerce-focused agents from malicious bots while maintaining customer relationships.
Delegation tokens extend OAuth 2.0 to anchor AI agent actions to human principals Personal Agent Protocol governs access control, not agent incentive conflicts Payments brand attachment narrows the AI payment trust gap significantly NIST proposes reusing existing identity standards for agentic systems Agentic commerce shifts fraud liability to merchants as fraud signals vanish
Does Personal Agent Protocol solve the identity and authorization gap for shopping agents? What identity and authorization standard would satisfy both consumer and merchant requirements? Can authenticated delegation tokens prevent merchants from absorbing agent payment liability? Are merchants actually adopting these protocols as an integrated stack? How does agentic commerce on crypto differ from consumer-authorized AI purchases? How does merchant-of-record status persist through API-based agent integrations? What payment methods does AP2 support beyond traditional cards? Should payments infrastructure adopt agent-specific protocols or reuse existing identity standards? Can a single protocol work across multiple AI agents and payment providers? Why do consumers trust Visa more than AI agents with payments? What trust gap does Visa aim to address with this protocol? How does Visa's protocol differ from OAuth-based delegation approaches?See all 15 inquiring lines on this note →
Will agentic commerce platforms adopt open or closed models?
Explores whether AI shopping agents will operate across multiple platforms or be locked into proprietary systems, and what structural consequences each path creates for competition and consumer choice.
Consumers expect AI agents to reshape purchasing but infrastructure gaps remain Platforms want agentic search before consumers rationally should switch Personal Agent Protocol governs access control, not agent incentive conflicts Authentication, liability, and fraud are the core barriers to scaling agent-delegated commerce Rothschild and colleagues argue standardization choice determines agent economy openness
Can agent platforms profit more from search than consumers benefit from it? Does Personal Agent Protocol solve the identity and authorization gap for shopping agents? Will agent platforms replace user attention as the economy's scarce resource? Can sellers influence their visibility through caching layer mechanics? What consumer and merchant benefits does agentic commerce create? What specific concerns prevent consumers from granting AI agents complete purchase autonomy? How would becoming an agent of choice protect brands in a multivendor basket world? Can retailers use off-site agents to compete without losing customer loyalty? What happens to traditional retailer-shopper relationships as third-party agents grow? Why do shoppers trust retailer-built agents more than third-party AI agents? Why would a platform prefer to control basket assembly rather than just product search? Which retail categories are most vulnerable to agent-driven basket fragmentation?See all 76 inquiring lines on this note →
What blocks autonomous agents from handling purchases at scale?
Federal Reserve Governor Waller identifies authentication, liability, and fraud as concrete barriers preventing AI agents from reliably making purchases on behalf of consumers. Understanding these challenges shapes how the payments industry builds trust in delegated commerce.
Personal Agent Protocol governs access control, not agent incentive conflicts Agent infrastructure must provide attribution, interaction, and response functions NIST proposes reusing existing identity standards for agentic systems Agent commerce exposes structural gaps in payments infrastructure designed for humans Agentic commerce shifts fraud liability to merchants as fraud signals vanish
Should retailers be able to refuse agents before any commercial relationship begins? What specific concerns prevent consumers from granting AI agents complete purchase autonomy? Can authenticated delegation tokens prevent merchants from absorbing agent payment liability? How might card networks recognize agent-operator liability to shift risk away from merchants? What fraud losses result when behavioral biometrics disappear in agent transactions? How do authorization boundaries protect users from unintended agent purchases? What enforcement mechanisms would make AI fiduciary duties work across jurisdictions? Why do most people distrust AI agents with payments? Could single concentrated banks face SVB-style runs from agents? Do uninsured depositors currently have agent access to move money faster? What liability risks emerge when sub-agents create spending authority autonomously? How do card controls like single-use limits address agent payment risk?See all 15 inquiring lines on this note →
Could federal AI agent registries solve platform access unfairness?
Should the FTC create a registry of trusted AI agents to ensure dominant platforms cannot exclude competing agents and restrict consumer choice? This explores whether certification and mandatory access rights can address market-concentration concerns.
Should AI agents serve one user or balance many interests? How can we monitor AI agents without enabling mass surveillance? Can existing identity standards work for AI agents? Can EU consumer law protect AI agents as algorithmic buyers? Does identity binding for AI agents chill whistleblowing and dissent?
AI alignment must expand beyond single-user loyalty to inclusivity and accountability Agent visibility requires identifiers, monitoring, and logs with privacy safeguards NIST proposes reusing existing identity standards for agentic systems EU consumer law's human-centric design fails for algorithmic consumer agents Identity binding for agents requires proportionate confidentiality protections by use case
Should retailers be able to refuse agents before any commercial relationship begins? What identity and authorization standard would satisfy both consumer and merchant requirements? Why does Amazon's objection focus on visibility and consent rather than credential safety? Can courts decide whether delegating your shopping to an AI agent is legally permissible? Could a cross-industry standard let users control what access agents have? How would becoming an agent of choice protect brands in a multivendor basket world? Can existing unfair commercial practices law cover agentic transactions today? How do existing consumer protection laws handle algorithmic agents as decision-makers? Can existing consumer protection laws address algorithmic shopping agents? What is the regulatory gap in Google's personalized pricing rollout? Can AI layers on TMC platforms change underlying OTA supply relationships? Does existing consumer protection law already provide tools for AI agent oversight?See all 19 inquiring lines on this note →