A Moment to Think · The argument before the cases

Marketing was a workaround

Marketing was the matching workaround for a human cognitive constraint.
The constraint has been lifted.

Marketing has one job, and has had it for over a century. Match a consumer's need with a brand's value proposition. Everything the marketing discipline built around that job (segmentation, positioning, brand, advertising, the funnel, attribution, performance, MarTech) has been an engineering response to a single problem. Humans could not do their own matching at the scale a market required. Marketers did the matching on their behalf, inside the cognitive constraints humans operated within. For a century, the discipline worked. The customer's agent now does the matching directly. The cognitive constraint has been lifted. The proxy is no longer required. The series this piece opens is an argument about what comes next.

The matching workaround

The matching workaround was real engineering. In a world of dozens, then hundreds, then thousands of choices in every category, no human could track and compare every offering against their needs. Marketers built solutions. Each of the classical marketing techniques was engineered around a specific human cognitive limit. No exceptions.

The agent removes the constraint

The customer's agent does what the workarounds approximated. It reads the customer's need from history, context, stated preference and live conversation. It reads each brand's value proposition from claims, audit trails, verified outcomes and structured attributes. It produces the match in real time, on the customer's terms, with verifiable reasoning. No segment, no positioning, no brand shortcut, no funnel stages, no click. None of the workarounds is required. ChatGPT, Claude, Gemini and Perplexity now cross half a billion weekly active users between them. The agent is not an experimental future. It is the working present. What marketers spent a century constructing workarounds to approximate, the agent now produces as a primary function of its existence.

Marketers were never the matchers. They were the workaround for the customer not being able to match. The customer just got the ability to do it directly. Marketing has been engineering itself out of a job for years and did not notice.

The disintermediation observation has been made before. Doc Searls predicted it in The Intention Economy (2012) and in the Cluetrain Manifesto (1999). Rishad Tobaccowala, Benedict Evans, McKinsey, BCG, Gartner and Forrester have published versions of it within the past two years. What this series adds is the underlying mechanism: the reduction of the discipline to one engineering constraint (Simon's bounded rationality, at market scale), and the consequence for what survives the lifting of that constraint.

What marketing has to become

The discipline that did the matching has to be rebuilt as the discipline that feeds it. Three jobs are central. Producing verifiable claims, so the brand the agent retrieves is the brand that survives the verification. Producing audit-grade decision logs, so every brand interaction with an agent leaves a record the agent reads as reputation and the regulator reads as evidence. Operator-controlled delivery, so the brand's customer record stays with the operator and the brand can change platforms without losing the record. Operators producing audit-grade decision logs under regulatory pressure for years (banks under SR 11-7, insurers under Solvency II, regulated service providers under conduct rules) have been doing this new marketing for a decade without realising they were the early adopters.

Marketing has always had a second job in parallel with matching: demand creation. The discipline of telling the customer about needs they did not know they had, categories that did not exist, products that solved problems the customer had not framed. Levitt named this in Marketing Myopia in 1960. Agents are bad at it. They match what the customer already wants against what the brand already produces. They do not invent new categories. New categories, new needs and new framing belong to the marketer, with the budget and the bandwidth freed up by no longer having to engineer the matching workaround.

Marketing was a workaround. The workaround is closed. What replaces marketing is the discipline of producing what the agent the workaround was substituting for now does directly.

What the series examines

The argument that follows works through what the matching migration produces in each part of the discipline. Seven themes.

The customer side

The customer is now represented by an agent that does not get tired and is not charmed. The marketing function now talks to counsel, not to the customer directly.

The buying side

Buyers have agents too. The buying committee, now five functions deep (CMO, CIO, CFO, CRO, COO), uses agents to evaluate vendor claims against operator data. The vendor demo selling on dazzle is selling to a room where half the evaluators are agents, and the agents are not impressed.

The operating model

Every executive's function is being redrawn around the agent layer that touches it. The CMO is now one of five at the table; the CFO, CIO, CRO and COO are the others. The org chart you end up with is the agent tech you bought.

The channel

Agent commerce is regulated by construction. Search regulated the claim; agent commerce regulates the decision. Five major regulatory anchors apply within an eighteen-month window.

Brand

Brand has migrated from the creative function to the decisioning function. The brand the agent reads is the audit trail the brand produced, not the story the brand told.

Measurement

Performance marketing's central unit, the click, was a proxy for human behaviour. The agent does not click. Performance is being rebuilt around the decision, not the click; around recorded cause, not reconstructed attribution.

Category

The MarTech category, as the industry has known it for twenty-five years, dissolves. The successor is architectural, not categorical. The fifty-billion-dollar budget is migrating to platforms that have a different shape.

Where the argument is weaker

The framing has limits worth naming. It is strongest for low-involvement repeat purchase, where the agent does the matching cleanly. It is weaker for identity goods, social signalling and high-involvement decisions where the buying motive is not "match my need." What the agent matches against will be gamed the way search results were gamed; adversarial optimisation for the model is the same engineering problem as adversarial optimisation for the click, and the discipline that wins is expert in both. And the framing assumes the agent is customer-owned. A gatekeeper-owned agent (Amazon, Google, Apple) re-intermediates the attention economy with a different toll. The optimistic case is the customer-owned agent. The cautious case is to design for both.

How to read what follows

Each piece in the series takes one of these themes and works it through. The pieces can be read individually or in sequence. The arc closes with a synthesis piece that pulls the seven faces back into one shape. The reader who finishes will have a working map of how marketing changes in an era where the customer can do their own matching, and the implications for the CMO, the CIO, the CFO, the CRO and the COO whose decisions about platforms will shape the function for the rest of the agent cycle. The constraint that built marketing has been lifted. The discipline that built the workaround has to be rebuilt around what the workaround was hiding all along.