Agents - and their better half: Part 1 - From agentic participants to agentic markets

Source: Sangeet Paul Choudary; Platforms, AI, and the Economics of BigTech (Substack) · 2026-10-04

Over the past few weeks, a new class of solutions is taking shape: a market-facing representative of the consumer.

Muse and Instinct - consumer-side AI agents - can search, evaluate and eventually transact on your behalf.

This seems like the natural endpoint of the agent economy: every consumer gets an agent. Instead of visiting multiple sites, they work through an agent.

But that framing may be too narrow. Super-apps and aggregators centralize the interface through which consumers access an otherwise familiar market.

Agents have the potential to do something far more transformative: not merely aggregate access to existing offers, but represent consumer intent in a form that can interact directly with agents on the other side of the market.

Markets are not made up of consumers alone. If agents are going to change markets rather than simply change how consumers navigate them, we need to look at what happens on the other side of the transaction.

The distinction is between a new locus of aggregation and a new mechanism of coordination. The former changes where the consumer enters the market. The latter can change the very structure of the market.

The promise of Muse and Instinct isn’t simply better search or evaluation of options - it is primarily representation in the market. The agent knows enough about an individual to act on that individual’s behalf.

But there is an inherent asymmetry in this view.

Markets are relationships between economic actors, while most agents today represent only one actor in that relationship. Muse may develop a rich understanding of what a consumer wants, but it largely encounters merchants through catalogues, product feeds, websites and APIs. A travel agent may understand a traveller’s preferences in considerable detail, but airlines and hotels still largely expose predetermined inventory, prices, and packages. A procurement agent may become considerably better at specifying a company’s requirements and evaluating suppliers, while suppliers continue to respond through conventional catalogues, quotations and bids.

The result is a lopsided agent economy.

Intelligence is accumulating on one side of many market relationships without an equivalent change in how the other side participates. This is easy to overlook because we tend to evaluate agents individually: how well an agent understands a user, reasons about alternatives or completes a task.

Markets, however, should be evaluated in terms of transactions/interactions. An agent’s capabilities are important, but what is equally important to executing the transaction is what happens when the counterparty to which it directs its intent can understand, reason and respond in kind.

Indeed’s Apply For Me allows an agent to act on behalf of a job seeker, while its Sourcing Assistant acts from the employer side by continuously identifying and engaging candidates. Indeed temporarily paused the automatic application mode of Apply For Me after employer feedback, noting that changing how applications are submitted also required new ways for job seekers and employers to engage with AI.

A candidate-side agent can know that someone is willing to accept a lower title for greater scope, or would relocate only for a sufficiently attractive opportunity, or has adjacent experience that makes an apparently imperfect match attractive. A conventional job description cannot reason about that intent.

Conversely, an employer-side agent can know which requirements are genuinely non-negotiable, whether compensation can stretch for an exceptional candidate, and which characteristics have predicted success in the role.

Once both sides are represented, hiring begins to look less like search against two static documents - a résumé and a job description - and more like interaction between two sets of objectives and constraints.

Imagine a traveller’s agent that knows that she is willing to leave a day earlier for a nonstop itinerary, strongly values a late hotel checkout, prefers one hotel chain but will switch for a substantially better room, and will spend more if a disruption can be handled flexibly. The agent can continuously search against this richer set of constraints.

But if airlines and hotels remain represented only by published fares, room categories and availability feeds, most of that intelligence cannot be applied. The traveller’s agent can optimize among pre-existing offers, but it cannot negotiate with the supplier to reconsider what they are willing to offer.

Now give the airline or hotel an agent as well. It could reason over inventory, yield targets, loyalty status, expected occupancy, ancillary capacity and the traveller’s expressed flexibility and respond with alternate offers that meet the buyer agent’s needs while meeting the seller’s constraints. The transaction is no longer produced by a buyer agent choosing from a catalogue. It is produced through mutual negotiation between represented parties.

A procurement agent might know that a manufacturer needs 5,000 units of a component, but also that delivery reliability matters more than the lowest unit price and that a certain technical specification can be relaxed. It can search far more intelligently than a conventional sourcing process.

Yet if suppliers continue to appear only as catalogues, price lists and responses to predefined RFQs, the buyer’s richer representation once again encounters a relatively static counterparty. The procurement agent becomes better at selecting from what suppliers have chosen to expose; the market itself does not become much more adaptive.

It could analyze production capacity, inventory, margins, alternative specifications, and the strategic value of the customer. It is representing the supplier’s ability and willingness to adapt.

We have agentic participants but don’t have agentic markets yet. More on that follows.

With an agent on only one side, intelligence improves navigation of the existing market.