Headless Commerce and the Death of the Funnel: What STAR Predicts

The consideration set used to form before the website visit. Now the purchase happens there too.

The funnel is dying. Not gradually, not in the way that marketers have been predicting and then un-predicting for fifteen years. It is dying because the thing that was supposed to be at the bottom of it — the purchase — has been decoupled from everything that was supposed to happen above it.

Berkeley’s California Management Review published a piece in April that should have generated more noise than it did. The argument, in essence, is this: AI agents are emerging as a new intermediary layer between firms and their customers. These are autonomous systems capable of searching, reasoning, and transacting on behalf of users. They filter information, synthesise alternatives, and execute transactions. The firm never gets the website visit. The consumer never sees the ad. The funnel collapses into a single, delegated decision.

This is not a prediction about the future. OpenAI’s Operator, Amazon’s Rufus, and Google’s Shopping agents are already executing transactions on behalf of millions of users. The consumer asks for running shoes for flat feet, under £150, suitable for trail. The agent compares across merchants, checks reviews, confirms availability, and completes the purchase. The consumer never visits a brand’s website, scrolls a product page, or consciously chooses between competing options.

If you work in marketing, this should make you deeply uncomfortable.

The Invisible First Click, extended

I wrote about the Invisible First Click framework earlier this year, arguing that in an AI-mediated environment, the consideration set is formed before the prospective customer ever visits a website. The AI assistant has already made recommendations, established credibility hierarchies, and effectively pre-selected a shortlist. By the time someone lands on your site, the decision has already been narrowed.

Headless commerce takes this further. The Invisible First Click said the consideration set forms before the visit. Headless commerce says the purchase can happen before the visit too. The entire funnel — awareness, consideration, evaluation, conversion — collapses into a single interaction between the consumer’s AI agent and the firm’s API.

The Berkeley CMR piece puts it bluntly: “If products are not machine-readable, they can be invisible to AI agents.” That is not a technology problem. That is a survival problem. If your product data, your pricing, your reviews, and your availability are not structured for machine consumption, you are not losing traffic. You are simply not present at the moment that matters most.

Which STAR types adapt fastest?

This is where the psychology gets interesting. The collapse of the funnel does not affect all consumers equally. It does not even affect all marketers equally. The speed and quality of adaptation depends on how people are wired, and the STAR Framework, built on Self-Determination Theory, Cognitive Bias Theory, and Regulatory Focus Theory, predicts the pattern with uncomfortable clarity.

Thinkers: The first movers

Thinkers are driven by competence (SDT). They need to understand the mechanism, evaluate the evidence, and make informed decisions. Headless commerce, paradoxically, serves them well. An AI agent that compares specifications across twenty merchants in three seconds is a Thinker’s dream tool. The agent does the cognitive labour the Thinker would have done themselves, but faster and at greater scale.

Thinkers will adapt to agentic commerce quickly because the tool aligns with their processing style. They will also be the first to notice when the agent’s recommendation logic is flawed, and the first to build workarounds. A Thinker who discovers that their AI assistant has a bias toward a particular merchant will switch assistants. They treat the agent as a tool, not a trusted advisor.

The risk for Thinkers is analysis paralysis at the system level. They will spend time optimising which agent to use, which preferences to set, which data sources to trust. The purchase decision is faster; the meta-decision about how to purchase becomes the new bottleneck.

Realists: The cautious adopters

Realists are driven by security (SDT). They need certainty, reliability, and the confidence that nothing will go wrong. Headless commerce triggers both their prevention focus (Regulatory Focus Theory) and their status quo bias (Cognitive Bias Theory).

A Realist’s first instinct is to distrust a system that removes their direct control over the purchase decision. Delegating a transaction to an AI agent means trusting that the agent will make the same choice they would have made, and that is a difficult leap for someone whose default orientation is toward risk mitigation.

But here is the thing about Realists: once they trust the system, they are loyal to it. The Realist who has their AI agent set up correctly, with clear preferences, verified merchants, and reliable return policies, will not switch. The status quo bias that made them slow to adopt becomes the same bias that makes them impossible to lose. The firm that earns a Realist’s agent-level trust has a customer for life.

Realists will be the slowest to adopt, but the most valuable to retain.

Adventurers: The ones who break it

Adventurers are driven by autonomy (SDT). They need freedom, novelty, and the sense that they are choosing their own path. Headless commerce is, for them, a threat.

The entire point of agentic commerce is delegation. The consumer sets preferences and lets the agent decide. For an Adventurer, that is not convenience. That is constraint. Their regulatory orientation is promotion-focused (Regulatory Focus Theory): they are drawn toward new experiences, new products, new possibilities. An agent that optimises for efficiency and price is optimising against everything the Adventurer values.

Adventurers will resist agentic commerce the longest. They will use AI agents for research but insist on making the final decision themselves. They will browse the product pages the agent skipped. They will choose the less efficient option because it feels more interesting. They will, in short, be the last humans actually visiting websites.

This makes Adventurers disproportionately important to brands that still rely on the display layer. If the funnel is collapsing for everyone else, the Adventurers are the ones still walking through it. Marketing to them is not about optimising for agents. It is about making the human experience compelling enough to override the convenience of delegation.

Socialisers: The ones who need convincing

Socialisers are driven by relatedness (SDT). They need connection, belonging, and the sense that their choices are validated by people they trust. Headless commerce removes the social layer from the purchase decision, and that is a problem.

A Socialiser does not buy running shoes by comparing specifications. They buy running shoes because their friend recommended them, because the brand feels like their kind of people, because the review from someone who sounds like them made them feel understood. An AI agent that optimises for price and performance is missing the dimension that actually drives the Socialiser’s decision.

Socialisers will adopt agentic commerce if, and only if, the social signal is embedded in the agent’s logic. If the agent can say “three people in your running club bought these,” the Socialiser is in. If the agent says “these are the highest-rated trail shoes under £150,” the Socialiser will still want to read the reviews themselves.

The brands that figure out how to inject social proof into machine-readable data will win the Socialiser. The brands that treat product data as pure specification will lose them.

What this means for marketers

The funnel is not shrinking. It is being bypassed. The four stages that marketers have spent decades optimising — awareness, consideration, evaluation, conversion — are being compressed into a single, machine-mediated transaction. The consumer does not see your ad. Does not visit your website. Does not read your carefully crafted landing page. The AI agent reads your API, compares your data to the consumer’s preferences, and either includes you in the recommendation or it does not.

If your product data is not structured for machine consumption, you are invisible. If your pricing is not competitive in a real-time comparison, you are excluded. If your reviews are not aggregated in a format the agent can parse, you are unpersuasive.

But the STAR analysis adds something the technology conversation misses: the consumer is not a monolith. The Thinker will delegate to the agent and optimise the meta-system. The Realist will adopt slowly but stay forever. The Adventurer will resist and keep visiting your website. The Socialiser needs the human layer to be embedded in the machine layer.

The brands that treat agentic commerce as a single problem will fail. The brands that understand it as four different problems, shaped by four different psychological orientations, will build systems that work for all of them.

The funnel is dead. The question is whether you are building for the consumers who are still walking through it, or the agents who have already walked around it.

David Chadderton is the creator of the STAR Framework and the author of three books, including The STAR Framework: Rewriting the Rules of Consumer Engagement (NYC Big Book Award 2025), The STAR Operating System: Decode Mindset, Understand Motivation, Transform Human Behaviour, and Dear Algorithm, It’s Not Me, It’s You. He writes about human behaviour, AI, and the psychology of decision-making on The Unoptimised Human.