Who captures value when AI agents cut coordination costs?

AlmaCap Research argues that AI agents compress the cost of the coordination work that sits behind most transactions — "searching, comparing, negotiating, scheduling, switching and claiming" — and that "profits that depend on coordination being expensive, through friction, inertia and breakage, compress" as a result. On the supply side, it argues vendor agents let small businesses automate their own coordination layer, so that "minimum efficient scale falls, firm counts rise, and intermediaries that sell coordination are squeezed."

On the demand side, AlmaCap argues delegation "will climb a ladder category by category, starting with low-risk recurring purchases," and that consumers currently "are delegating research faster than they are delegating checkout" — a gap between agents doing the comparison work and agents doing the committing work. The combined effect, it argues, pushes markets "closer to the textbook": narrower price dispersion, negotiated prices at low values, more cross-border competition — the classic signature of falling search costs. But it pairs that prediction with a structural caveat: a "new concentration risk at the agent layer," because value migrates not only to sellers with "unique supply, physical capacity, trust and payment rails" but also to "the agent platforms" themselves, which sit in front of every transaction they mediate.

This is a market-structure argument built on the same coordination-cost vocabulary that How do coordination architectures trade off efficiency against resilience? formalizes for abstract task assignment: AlmaCap's claim that falling coordination costs squeeze intermediaries echoes the older models' finding that centralized schemes lower coordination cost at the price of other costs elsewhere, here recast as a question of who captures the savings rather than which topology is more efficient. It also names the same winner as Will agents compete for attention just like users do?: both put the agent platform, not the seller or the buyer, at the center of the post-delegation economy. And its concentration worry has a formal parallel in Why do consumers delay adopting agentic search?, which shows a platform's revenue incentive and a consumer's welfare can diverge even within a single agentic-search model — one mechanism by which AlmaCap's "handful of agent operators" could capture surplus the textbook prediction assumes would reach consumers.

As a paid Substack's free-preview summary, the piece states predictions and a framework, not measurements — there is no cited data on actual price dispersion, firm entry, or agent-platform margins, and the question of who captures the surplus is presented by AlmaCap as open, not answered. The implication it draws is conditional: if delegation proceeds as described, the textbook gains from lower search costs are real but not guaranteed to reach consumers, because the same delegation that disciplines seller margins also hands a new bottleneck — the agent layer — pricing power over the transaction.