Wiki › Economy, compute and governance

Agent-Mediated Markets

Consumer markets where personal agents research and buy for people, weakening firms' edge from bounded rationality and shifting rewards toward product quality.

Agent-mediated markets are consumer markets in which purchases are researched, compared, and increasingly executed by Personal Agents acting for the buyer. The core hypothesis, set out by Patrick Collison, is that agents shrink the rationality gap between firms and consumers — and that this changes which business strategies pay off.¹

Firms vs. bounded rationality

Collison frames companies as "the original superintelligence": they pool cognitive effort across millions of transactions and can optimize against consumers who are inattentive, underinformed, or irrational. Herbert Simon's *bounded rationality* is often a source of firm profit — the classic example is an insurance upsell at the end of a booking flow, priced well above a standalone policy. A personal agent that never gets tired or distracted narrows that gap.

What shifts

• Price discrimination weakens. Coupons and similar screening devices only work if some buyers don't bother; an agent that always hunts for discounts breaks the mechanism.
• Cross-subsidies change. Today people who forget to cancel subscriptions subsidize those who remember. Agents that track and cancel erode such models. The aggregate incidence across the economy is unclear.
• Quality beats distribution, on the margin. If every buyer behaves as if they spent ten hours researching, "everyone already buys ACME mousetraps" matters less and a genuinely better product matters more. Agents act as a structural subsidy for product quality.
• Quality signals gain value. Agents have an insatiable appetite for reviews, test data, and other quality evidence. Data becomes more valuable not only as model-training input but as a complement to cheap intelligence.

Open questions

• Superstars or dispersion? Rational agents might converge on the same "best" product (stronger winner-take-all), or, by eliciting individual preferences, break the artificial clustering created by mass marketing and widen variety.
• Demand routing. Many internet businesses monetize routing short-attention demand — ranking, placement fees, first-result clicks. Some of this is a real signal (willingness to pay), much is deadweight. How its value changes when the "clicker" is an agent is unknown but potentially large. This is the buyer-side mirror of the same shift seen in Agentic AI in Retail, where sellers also deploy agents.
• Whose side is the agent on? The whole argument assumes consumer-aligned agents. Collison expects that to be the winning strategy but notes it raises many further questions — an agent monetized by sellers would reverse most effects above.

Synthesis

Collison describes the market as "a massively parallelized, decentralized, and highly imperfect value attribution machine." His bet is that agents change its backpropagation so that rewards for better products increase. This is a single-source, explicitly speculative view; it sits in tension with darker scenarios such as Dead Economy Theory, which focus on income and demand rather than on the quality of consumer choice.