For thirty years, I've watched our industry get extraordinarily good at one thing: understanding the human mind at the moment of purchase.
We can map how awareness becomes association, how association becomes preference, how preference becomes loyalty. We can measure mental availability and correlate it with sales. We can engineer the fast, emotional System 1 "yes" that drives most FMCG decisions. Aaker, Keller, Sharp, Kahneman — between them, they gave us a near-complete science of the human buyer.
And every bit of it assumes the buyer is human.
I've come to believe that assumption is the most important — and most overlooked — fault line in modern brand strategy. Because the buyer is changing.
The decision-maker is leaving the room
When a consumer delegates grocery replenishment to Amazon Alexa+, or asks Apple Intelligence to find them a good option, something quietly profound happens. The consumer doesn't disappear. But they stop selecting. An AI agent makes the selection on their behalf.
Stephan Puntoni captured this in Harvard Business Review this year: AI is disrupting marketing on two fronts — as a tool that changes how we work, and as a consumer that changes who decides. The second front is the one that should reorganise our priorities. Because the agent is not a smaller, faster human. It is a fundamentally different kind of evaluator.
It doesn't feel. It doesn't remember. It doesn't recognise your pack or hum your jingle. As Byron Sharp taught us, a human brain thinks of brands — mental availability is the probability of being recalled in a buying situation. An agent doesn't think of brands. It retrieves them. It processes structured attributes, weighs ratings and reviews, considers endorsements and certifications, and selects. The emotional resonance we've spent decades building has no direct point of entry into that logic.
What the evidence already suggests
This isn't speculation. A growing body of published research is mapping how agents actually choose, and the early signals are striking — and consistent.
In a randomised experiment across eight product categories, an "Overall Pick"-style endorsement badge raised an agent's selection probability from a roughly 10% baseline to somewhere between about 20% and 43%. Meanwhile, sponsored and paid-placement tags pushed selection *below* baseline. Paid prominence — a cornerstone of digital-shelf strategy — was penalised, not rewarded.
In separate work spanning thousands of simulated shopping rounds, traditional persuasion tactics — scarcity, urgency, strikethrough pricing, bundle framing — proved inconsistent or counterproductive. Advanced reasoning models appeared to interpret overt persuasion as a signal of low quality. The signals that survived agent scrutiny were the ones you can't manufacture: authentic ratings, genuine review volume, honest claims, third-party certification.
I want to be disciplined about what this proves. It's early. Agent behaviour is unstable across model versions — a single update can shift the picture, and some of the more dramatic figures circulating still need peer-reviewed replication. Anyone claiming certainty here is overreaching.
But the direction is hard to dismiss: agents reward verifiable truth and earned authority, and distrust manufactured persuasion. That is close to the inverse of the incentives many brands have optimised for.
The missing word
Here's what strikes me most. We have an exquisite vocabulary for the human front — equity, salience, resonance, availability. We have almost no vocabulary for the other front. And you cannot manage what you can't name.
So let me offer a working term: agent legibility — the degree to which a brand's equity signals are interpretable, accessible, and positively weighted by AI purchasing agents.
The key is where it lives. Human brand equity is a property of consumers' minds. Agent legibility is a property of your brand's information environment: the completeness of your structured product data, the quality and volume of your authentic reviews, your certification and endorsement profile, the absence of the manipulative markers advanced agents penalise, and your prominence in the data environments agents actually query.
Notice that fourth dimension. For a human, a bold claim can persuade. For an agent, the absence of overt persuasion can be the asset. That single inversion should tell you how different this game is.
This complements brand-building. It doesn't replace it.
I want to be very clear, because this is where the conversation usually goes wrong: I am not arguing that brand-building is dead, or that emotion stops mattering.
The opposite, in fact. Human equity changes its job in the agentic era — it becomes the reason a consumer names your brand when instructing an agent, the reason they override a default, the standing preference the agent executes. A brand people love is a brand people tell agents to buy. That role isn't smaller. It may be more strategic.
Agent legibility sits alongside the traditional dimensions, not on top of them. Together, they form what I think of as a two-front model of brand equity. Front One wins the heart. Front Two earns the machine. And the new core competence is orchestrating investment across both.
Why I'm naming it now
I'm not presenting agent legibility as a finished theory. It's a working construct, and we're about to research it properly — through in-depth conversations with senior FMCG
brand leaders, to test whether it holds, discover its true dimensions, and understand how practitioners are actually navigating both fronts at once.
But even as a working idea, naming it has already changed the conversations I'm having. Once a leadership team can say out loud, "We're strong on human equity and
weak on agent legibility," they can finally do something about it. The naming is the first act of management.
So I'll leave you with the question I keep asking myself: your brand scorecard measures the first half of equity beautifully. Does it have a single line for the second?
If it doesn't yet, that's not a gap in your reporting. It's a gap in the map.
