The Market Becomes an Interface

How Allocation Moves Upstream Before Buyers and Sellers Ever Meet

The market does not disappear. It becomes the part of the system the user is still allowed to see.

The buyer still chooses. The agent decides what can be chosen. And before either acts, the pre-market has already decided who is eligible to compete.

The market does not need to disappear to stop behaving like the market people imagine.

Prices can remain. Buyers and sellers can remain. Jobs, products, services, contracts, ads, subscriptions, stores, banks, platforms, and consumers can all remain. The surface vocabulary of market society can survive almost intact.

And still, the decisive act of allocation may move somewhere else.

The old market was imagined as a place where actors met. Buyers met sellers. Workers met employers. Borrowers met lenders. Producers met consumers. Prices coordinated the exchange. Competition disciplined the participants. Information was imperfect, power was unequal, and access was never as open as theory suggested. But the moral image remained one of encounter.

That image is becoming obsolete.

The new market is not only a place where actors meet. It is a system that decides whether they can meet at all.

Search decides what can be found. Ranking decides what can be seen. Identity decides who can be trusted. Payment rails decide who can be monetized. Compliance decides who can participate. Procurement decides what institutions are allowed to buy. Agents decide what is worth comparing before a person ever sees the field.

The market remains visible.

Allocation moves upstream.

The market becomes an interface.

The Pre-Market: Where Allocation Actually Happens

Standard accounts of market failure ask what happens inside the market: monopoly, price distortion, information asymmetry, externalities, unfair competition. But that assumes the market is still the operative arena.

Increasingly, the decisive layer sits before the arena.

What is emerging is a layer that operates before the market acts. Call it the pre-market: the system of identity verification, eligibility classification, compliance filtering, payment authorization, procurement scoring, ranking logic, and agentic selection that determines who enters the market before any transaction begins.

The pre-market is not the market. It does not look like a market. It is not regulated as a market. It presents itself as neutral infrastructure: trust systems, safety checks, identity rails, efficiency tools.

But it does more allocative work than the visible market it precedes.

A seller excluded at the pre-market layer is not outcompeted. She is never entered into the contest.

This is where the platform-economy literature needs an additional layer. Existing accounts describe how arenas are distorted: how ranking changes competition, how gatekeepers extract rents, how network effects entrench dominance, how platforms mediate demand. Those analyses are right. But they do not fully capture the threshold before the arena.

The pre-market is that threshold.

Eligibility comes before price. That is the key move.

In classical market theory, price is the sorting mechanism. You compete by offering a better deal. In interface markets, you must first be eligible to make an offer at all. The seller with the better product who fails the agent’s eligibility filter does not lose on price.

She never appears.

A market organized around visible price competition is a different institution from a market organized around machine-legible eligibility. The first asks actors to win in public. The second asks actors to satisfy criteria embedded inside systems they may not see, contest, or even know exist.

That distinction is not a refinement of market theory.

It is a structural transformation of what markets are.

The Platform Was Only the First Interface

Platform capitalism made this structure visible before AI.

A person opens an app to choose a restaurant, hotel, product, article, investment, route, video, doctor, or flight. The experience feels like choice because there are options. The user scrolls, compares, filters, clicks, and buys.

But the field has already been shaped.

Search decided what appears. Ranking decided what appears first. Reviews converted experience into scores. Ads purchased placement. Personalization narrowed the world. The platform set the categories.

The user sees a market.

The system sees a ranked field.

That difference matters because ranking is not merely presentation. Ranking is allocation. To appear first is to receive opportunity. To appear late is to be technically available but practically absent. To be excluded from the recommendation layer is to exist without circulation.

The same structure appears from the seller’s side. A seller may own the product but rent the demand. Traffic comes through search, social feeds, marketplaces, ad auctions, payment processors, and platform rules. A change in ranking can alter revenue overnight. A policy update can remove an entire category.

Amazon’s marketplace is the clearest public example. The Federal Trade Commission’s antitrust case against Amazon alleges that Amazon uses interlocking marketplace practices affecting sellers, fees, search, advertising, and competition in ways that shape seller access to demand. Amazon disputes the allegations. But the case matters because it names the structural issue: a marketplace can become the environment in which sellers must operate while remaining formally independent. [1]

The European Union’s Digital Markets Act designates large platforms that provide core platform services and occupy positions through which other firms must often pass to reach users. The point is not merely that these firms are large. The point is that they sit between actors who still believe they are participating in markets. [2]

A seller can spend years building demand, only to discover that demand was never fully hers. It belonged partly to the ranking system, partly to the ad auction, partly to the platform’s definition of trust.

The dependency does not look like dependency.

It looks like a dashboard.

But platform interfaces, for all their power, still left the market partially visible. The user could scroll. The seller could measure rank. The interface mediated. It did not yet fully delegate.

AI pushes the market into a different phase.

The Interface Stops Being a Screen

A platform shows options.

An agent acts.

That is the key AI turn.

A search engine returns results. A feed orders content. A marketplace ranks products. A hiring platform filters résumés. These systems shape the field, but the user still sees something like a market: a list, a grid, a ranking, a menu, a feed, a shortlist. She can still imagine that she is choosing among options placed before her.

An agent changes the structure.

An agent can search, compare, summarize, negotiate, purchase, schedule, route, reject, recommend, and execute on behalf of the user. It does not merely display the market. It performs part of the market.

The interface stops being a screen and becomes a delegate.

This is a phase change, not an extension of what came before. The platform interface still left the market partially visible. Its power was enormous, but it appeared as mediation. The user could scroll. The seller could measure rank. The agentic interface is different because it can compress the entire comparison process into a completed selection.

The market no longer needs to appear as a field of options.

It can appear as an answer.

The old interface said: here are the choices.

The new interface says: here is what I chose for you.

The buyer still chooses, but choice moves to a higher level of abstraction. She chooses the agent, the instruction, the preference profile, the trust setting, the budget, or the desired outcome. The agent then chooses among the market’s offerings.

That does not eliminate human agency.

It relocates it.

McKinsey projects that AI agents could mediate between $3 trillion and $5 trillion of global consumer commerce by 2030. These forecasts are usually framed as retail innovation. The political-economic significance is larger. [3]

When agents mediate markets, competition shifts from persuading the buyer to being selected by the buyer’s delegate. The buyer may never see the full field. The seller may never know why it was excluded. The market is still operating somewhere behind the agent, but the human no longer experiences it as a market.

She experiences it as assistance.

The Buyer No Longer Sees the Whole Field

The human market was organized around visible comparison.

That visibility was never perfect. Advertising distorted attention. Shelf placement mattered. Geography mattered. Brand power mattered. Search costs mattered. But the market still had a human-legible form: a buyer saw options, compared prices, judged quality, and decided.

An agentic market does not need to preserve that form.

The agent can compare more options than the user would ever examine. It can weigh delivery speed, return policy, contract terms, trust score, compatibility, past behavior, fraud risk, sustainability tags, inventory status, warranty conditions, payment acceptance, user preference, institutional policy, and price at the same time.

This can be genuinely useful. It can save time. It can reduce search costs. It can protect users from bad offers. It can negotiate better terms. It can remember preferences. It can reduce cognitive burden.

That is why it will spread.

The interface does not conquer the market by force.

It becomes powerful because it is useful.

But usefulness is not neutrality.

Imagine a traveler who no longer compares twenty hotels. She tells the agent: book me somewhere quiet, under $250, near the conference, with late checkout and reliable Wi-Fi. The agent filters the market before she sees it. Hotels without structured cancellation data, verified reviews, live inventory, compatible payment rails, or trusted booking integration may never appear.

They are not rejected by the traveler.

They are excluded by the conditions of delegation.

Once the agent becomes the buyer’s delegate, the market must be organized for the delegate. Sellers no longer compete only for human attention. They compete to be selected by a system that may optimize across criteria they cannot fully see.

A hotel may be excluded because its cancellation policy is poorly formatted. A product may be ignored because its warranty is not machine-readable. A restaurant may disappear because its menu data is inconsistent. A contractor may be filtered out because its insurance credential is not verified through the right channel. A small seller may lose not because the buyer disliked the product, but because the agent did not consider the seller eligible for comparison.

The exclusion happens before persuasion.

The seller is not rejected by the buyer.

The seller is never presented to the buyer as a serious option.

The New Bottleneck Is Eligibility

In classic market language, the central question is often price.

Can the buyer afford it? Can the seller compete? Can the worker command a wage? Can the firm underbid a rival?

In interface markets, price still matters. But the pre-market comes first.

Before price matters, eligibility matters.

Can the actor be found? Can the identity be verified? Can the product be categorized? Can the vendor pass procurement? Can the agent parse the offer? Can the payment system tolerate the risk? Can the algorithm place the actor into circulation?

These are not secondary frictions. They are increasingly the conditions of market existence.

The firm that cannot pass procurement does not reach the institutional buyer. The worker filtered out by automated screening does not reach the interview. The creator demonetized by platform rules does not reach economic sustainability. The startup blocked by API access, cloud cost, or app-store rules does not reach scale.

The market remains formally open, but practical access is governed upstream.

Agentic markets intensify this because the buyer’s delegate may never display the arena at all. A seller who fails the agent’s eligibility layer is not merely ranked lower. She may be absent from the final answer. A vendor who fails machine-readable procurement criteria does not become less competitive. It may never enter the institutional choice set.

A product that cannot be parsed by the agent is not badly marketed in the ordinary sense.

It is commercially illegible.

The old question was: can you compete?

The new question is: can the system recognize you as eligible to compete?

That shift, from competition to eligibility as the operative sorting mechanism, is the central transformation the pre-market introduces. It cannot be captured by existing antitrust frameworks alone, because antitrust asks whether competitors are treated fairly inside a market. The pre-market problem is that the excluded actor never reaches the market in the first place.

This is why the pre-market is politically harder to name than platform power. Platform dominance is visible: you can measure rank, compare search results, audit recommendation systems, document seller fees, identify gatekeepers. Pre-market exclusion often produces no signal at all.

The excluded actor does not know she was excluded.

She receives silence where she expected market feedback.

No rejection.

No ranking.

No price signal.

Just absence.

The Market Becomes Machine-Legible

The interface market does not merely allocate. It makes actors legible.

To participate, people and firms must format themselves in ways the system can rank, score, route, verify, approve, and compare. They need ratings, metadata, credentials, reviews, histories, compliance records, identity proofs, performance metrics, structured descriptions, and machine-readable claims about what they are.

In the old market, the buyer had to understand the seller.

In the platform market, the platform classified the seller.

In the agentic market, the buyer’s delegate must be able to parse the seller before the buyer ever appears.

The seller must become readable not only to people, but to systems acting for people. Product data must be structured. Reputation must be machine-readable. Policies must be interpretable. Trust must be formalized. Inventory must be accessible in real time.

The buyer’s agent will not browse like a person.

It will parse, score, compare, eliminate, and execute.

This produces a new kind of market discipline. The actor who refuses to become legible may preserve integrity but lose circulation. The actor who becomes legible gains access but becomes flatter, more standardized, more shaped by the categories of the system that reads her.

The restaurant optimizes for review prompts rather than the meal. The creator optimizes for retention curves rather than the argument. The worker optimizes the résumé for the tracking system rather than the role. The seller optimizes for search rank rather than durability. The doctor optimizes for platform-compatible availability rather than care continuity.

This is not merely gaming.

It is adaptation to the market’s new nervous system.

The interface does not merely read the market.

It teaches the market how to deform itself for readability.

The Agent Is Not Neutral

The agent will be sold as personal empowerment. That claim will often be true. A good agent can protect the user from manipulation, reduce decision fatigue, compare hidden fees, detect bad contracts, and negotiate better terms.

But every delegate carries a politics of selection.

The agent does not need to be corrupt to shape the market. It only needs criteria. And every criterion is an allocation rule.

Some criteria are explicit: price, speed, ratings, availability, compatibility. Others are embedded inside models, payment partnerships, risk systems, safety filters, and default settings the user never inspects. The distinction matters because the agent’s hidden architecture, including which merchants are integrated, which payment rails are supported, which vendors are considered safe, and which offers are machine-readable, determines the field before the user’s stated preference applies.

This is where payment infrastructure becomes political economy.

Visa’s Intelligent Commerce program describes payment infrastructure that lets AI agents integrate into global commerce with secure, reliable payment experiences. Mastercard’s Agent Pay presents agentic payments as a trust layer in which registered agents transact through governed, traceable network tokens. [4] [5]

Both are describing the same structural reality: for the agentic market to function, the delegate must be able to pay, not merely recommend.

Once the agent can pay, it stops being a search tool and becomes an economic actor.

Recommendation is still speech.

Payment is allocation.

At that point, the payment infrastructure beneath the agent, including its integrations, risk classifications, merchant eligibility rules, and fraud thresholds, is not plumbing. It is an access layer. The ability to receive money and the ability to pay are no longer neutral utilities. They become conditions of market entry.

A seller excluded by a payment system’s risk engine is not excluded by price. She is excluded by classification before any buyer forms a preference, before any agent runs a comparison, before any offer is weighed.

The seller’s audience changes. First she competed for attention, then for ranking. In the agentic market, she competes to enter the agent’s eligibility set before the buyer forms a preference at all.

That is a different market.

It requires a different political economy to describe it.

The State Arrives Late

The state is not absent from interface markets. It regulates privacy, competition, consumer protection, labor classification, payments, and platform conduct.

But it consistently arrives after the allocation system has already formed.

This lateness is structural, not accidental. Private interfaces scale through coordination before they become public infrastructure. A platform solves a problem, grows because it reduces friction, becomes habitual, then infrastructural. By the time regulators recognize its market power, firms, workers, consumers, and public agencies have already reorganized around it.

Regulation then threatens not merely the platform, but the dependencies built on top of it.

AI makes the classification problem harder. A platform can be audited as a platform. An agentic interface is more diffuse. It appears inside browsers, enterprise software, procurement systems, financial apps, healthcare triage systems, workplace automation suites, legal tools, and personal assistants.

It does not always look like a market gatekeeper.

It looks like productivity.

The pre-market is harder to regulate than the market because it does not look like the thing it is. Antitrust law looks for dominant firms, exclusionary conduct, and foreclosed competition. But upstream exclusion often works through fragmentation. No single system excludes everyone. Many systems together exclude specific actors through compliance requirements, identity standards, payment classifications, procurement rules, and machine-readable trust.

Each individual gate looks reasonable.

The aggregate effect is a system that determines participation before the visible economy acts.

The state is trained to see the transaction. The interface has already shaped the field before the transaction appears.

That is why the state arrives late even when it acts aggressively. It is not merely behind in time. It is behind in ontology. It comes looking for the market as a place where actors meet, while the market has become a system that decides whether they can meet at all.

Rented Intelligence Begins Here

This is where interface markets begin to produce a new class condition.

The new divide is not only between those who work and those who own. It is between those who own the interface and those who must become legible inside it.

This is the beginning of rented intelligence environments: economic spaces where actors appear independent, but rely on upstream systems for cognition, distribution, compliance, payments, trust, and customer access. The older platform economy rented distribution. The AI interface economy begins to rent cognition and distribution together.

A seller may still own the product. A creator may still own the voice. A worker may still own the skill. But the systems that make those assets visible, trusted, ranked, monetized, compared, and executable increasingly sit elsewhere. The formal language remains entrepreneurial. The practical condition moves toward tenancy.

The dependency is fragmented across systems: payments, ads, marketplaces, logistics, cloud services, model providers, app stores, identity layers, compliance tools, procurement systems, recommendation engines. That fragmentation makes it harder to see.

The actor is not commanded directly.

She is conditioned environmentally.

For now, it is enough to see the beginning of that condition: the market becomes an interface, and the interface becomes the environment in which economic action is allowed to appear.

Choice Survives as an Experience

The market does not end in this scenario. That is what makes the transformation hard to see.

The vocabulary of market society remains. Consumers, sellers, workers, creators, and firms all continue to perform their familiar roles. But each act increasingly passes through upstream systems that structure the field before the formal market act begins.

A society can preserve the rituals of market freedom while relocating the substance of market access. It can keep prices, choices, competition, entrepreneurship, and consumer abundance while the real struggle moves into ranking, permission, eligibility, legibility, payment, procurement, and agentic selection.

The market still feels like freedom because the interface still feels like a tool. The buyer experiences assistance. The seller experiences optimization. The worker experiences screening. The creator experiences analytics. The firm experiences workflow.

Each experience is local.

The structure is systemic.

Beneath those experiences, allocation has moved. The upstream system has already decided who gets seen, trusted, ranked, parsed, compared, selected, and executed before the market appears.

The market remains visible.

Allocation has already happened.

The pre-market has no waiting room.

You do not know you have been excluded.

You simply find that the market does not respond.

The buyer still chooses. The agent decides what can be chosen. And before either acts, the pre-market has already decided who is eligible to compete.

That is the political economy of the interface. The market does not disappear. It becomes the part of the system the user is still allowed to see.

Sources

[1] Federal Trade Commission, “FTC Sues Amazon for Illegally Maintaining Monopoly Power.” The FTC alleges that Amazon uses interlocking practices affecting sellers, fees, pricing, advertising, search, and competition. Amazon disputes the allegations. https://www.ftc.gov/news-events/news/press-releases/2023/09/ftc-sues-amazon-illegally-maintaining-monopoly-power

[2] European Commission, Digital Markets Act Gatekeepers Portal. The DMA designates large platforms as gatekeepers when they provide core platform services that function as important gateways between businesses and users. https://digital-markets-act.ec.europa.eu/gatekeepers-portal_en

[3] McKinsey, “Agentic Commerce: How AI Shopping Agents Can Change Retail.” McKinsey projects that AI agents could mediate $3 trillion to $5 trillion of global consumer commerce by 2030. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-automation-curve-in-agentic-commerce

[4] Visa, Intelligent Commerce. Visa describes infrastructure for AI agents, platforms, and developers to integrate into global commerce with secure and reliable payment experiences. https://corporate.visa.com/en/products/intelligent-commerce.html

[5] Mastercard, Agent Pay. Mastercard describes agentic payments using registered agents and governed, traceable network tokens. https://www.mastercard.com/us/en/business/artificial-intelligence/mastercard-agent-pay.html