Something has changed in the order of operations in commerce, and most marketing plans have not yet priced it in.
For a century, the discipline rested on one assumption: that the customer arrives at the moment of decision carrying doubt, and that our job is to resolve it in the aisle, at the shelf edge, on the last click. We built entire operating models on this. It is now being retired, not because customers have stopped deciding, but because they’ve started delegating.
Accenture’s study, Me, my brand and AI: The new world of consumer engagement, which surveyed roughly 18,000 consumers across 14 markets, puts numbers to the handover. Seventy-two per cent of consumers now use generative AI regularly. Around 36 per cent of active users describe it as a “good friend”. Nearly one in ten already rank it as their single most important source of purchase recommendation, second only to physical stores among active users. Seventy-five per cent say they are open to a trusted AI personal shopper.
The comfortable reading is that this is another distribution shift, the sort our industry absorbs every decade. Search changed how demand was found. Social changed how it was influenced. Retail media changed where it was harvested. This is different in kind. For the first time, the deciding itself is being outsourced to something with no memory of your Christmas campaign, no affection for your founder’s story and no interest in your brand book. An agent told to find the best option on price and specification will do precisely that, faithfully and at scale.
Accenture’s formulation of what survives the encounter is the report’s sharpest sentence: trust becomes the only currency AI cannot commoditise.
Compete for the brief, not the click
Here is the operating consequence, and it is the part the trend coverage keeps missing. If the machine runs the comparison, the marketer’s battleground moves upstream, not to the consideration set, which the agent now assembles, but to the instruction the customer types before the agent starts work.
There are only two kinds of instruction. “Find me the best noise-canceling headphones under two thousand” is a specification, and in that context your brand is a row in a table, adjacent to a cheaper row. “Order me the pair I tried in the shop” is a preference, and the search is already over. One is a commodity auction. The other is a brand asset expressing itself.
So every unit of brand investment now faces one clarifying test: does it raise the probability of being named, unprompted, in the customer’s instruction to their machine? That is not a philosophical question but a measurable one, and share of prompt will become as familiar to the boardroom over the next three years as share of voice was to the last generation of chief executives. It also settles an old argument.
Emotional loyalty was always the soft end of the budget, the part cut first when the quarter tightened. In an agent-mediated market, it is the only part that cannot be arbitraged away, because everything downstream of the instruction is now automated, and automation drives returns towards zero.
The human premium is an economics story, not a values story.
The second misreading is to treat human-centred marketing as a moral posture, a nicer way to do the same job. It is a scarcity argument, and it should be presented to a board as one.
Generative AI has collapsed the marginal cost of competent content to near zero. When competence becomes infinitely available, the premium migrates to whatever remains scarce: taste, judgement, a point of view, visible craft. The industry is already pricing this in.
HubSpot’s State of Marketing 2026 finds that 63 per cent of marketers say they need more unique, human-centred content to stand out, and 61 per cent agree that brand point of view and human taste matter more, not less, when humans and AI work together.
Ogilvy’s The Human Premium points in the same direction, identifying demand for intentional creative expression, visible craft and patina, niche communities and human curators acting as trusted filters against algorithmic overload.
Consumers are already onto this. In the Accenture data, 45 per cent say AI-generated content lacks personability and 41 per cent question its authenticity. Which explains why “made by humans” is emerging as a positioning signal of the same order as “handmade”: a claim about provenance rather than performance.
And that is where the next three years get uncomfortable, because provenance claims get audited. “Handmade” endured as a premium because it was verifiable. “Made by humans” will not survive as an assertion in a deck.
The directional shift is from claim to disclosure: named creators, published process, visible making, credits on the work. The brands that hold the premium will not tell you they are human. They will show the receipts, and any enterprise leaning on authenticity should assume it will one day have to prove it.
Personalisation has already failed once.
There is a cautionary data point that deserves more attention than it gets. VML’s Future Shopper report, covering more than 25,000 shoppers across 16 countries, finds that 45 per cent of consumers think brands do a poor job of personalisation, with recommendations judged irrelevant and emails generic.
Read that against the decade that produced it. Personalisation under-performed during the period of maximum data, maximum compute and maximum executive enthusiasm. More compute will not repair it, and any human-centred strategy built on better targeting rests on the one capability the market has already tested and marked down. The differentiator is not precision. It is relevance and participation, which is a different engineering problem and a different creative one.
The same report shows where the energy went instead. Fifty-nine per cent of shoppers are more likely to buy from companies that speak and act positively on diversity and social issues, and nearly half have shifted habits towards brands aligned with their political views. Read that with clear eyes rather than enthusiasm: values-led positioning is a sorting mechanism, and sorting wins customers and loses them in the same motion. It demands conviction, which is expensive, and it has to translate into tangible benefit rather than statements.
Dentsu’s trend work captures the paradox: the same consumers who have embraced AI are actively seeking offline, unplugged, handmade experiences.
What has to break inside the building.
None of this is achievable with the current machinery, which is where the enterprise question begins.
Ashley Faus makes the structural case in Human-Centered Marketing: How to Connect with Audiences in the Age of AI, drawing on work at Adobe, HubSpot, Edelman and Atlassian. Her core move is to replace the funnel with a playground: audiences enter and exit freely, consume content in any order and use it “wrong”. As a metaphor, it has been around for a while. As an operating model, it has suddenly become urgent, because the funnel was fundamentally a machine for harvesting intent, and intent harvesting is exactly what the agents now do better, cheaper and without us. What cannot be automated is manufacturing preference before intent exists.
That inverts the budget logic of the past ten years. Spend that captures demand is heading towards commodity pricing. Spend that creates demand is heading towards scarcity pricing. Faus’s other two prescriptions follow from the same reasoning: evolve social from broadcast into genuine community facilitation, and build thought leadership on credibility, profile, prolific output and depth of ideas, because those are the four things a machine cannot borrow on your behalf.
Three consequences for anyone rebuilding an operating model this year:
- First, optimise for two audiences at once: human discovery and machine visibility, which means structured data and generative and answer engine optimisation alongside conventional search. The point is not merely to be found but to be described accurately, because the machine now paraphrases your brand to your customer, and mushy positioning produces a mushy paraphrase.
- Second, move measurement off short-term conversion and onto trust, emotional attachment, retention and community health, the leading indicators of being named in a brief.
- Third, put AI on scale, analysis, variation and efficiency, and reserve the high-stakes emotional, creative, and relationship work for people. That is not sentiment. It is capital allocation.
Bravery as an operating discipline.
Which brings me to the word in the name. The material risk in this era is not rogue technology. It is convergence. Every competitor is buying similar models, training on similar data, chasing similar benchmarks and writing similar prompts, and the statistical mean is becoming the most crowded place in the market. Sameness is now the default output of a well-run enterprise.
Difference therefore has to be manufactured deliberately and defended structurally. Someone senior is accountable for what the organisation will not automate. A protected budget line exists for craft that attribution cannot justify. And there is tolerance for ideas that test poorly and travel far, because work that offends no algorithm rarely moves anyone.
The winning brand is not the company with the best AI stack. Within eighteen months, everyone will have a comparable one, and it will confer no advantage whatsoever.
The winning brand is the company that has decided, explicitly and at board level, which of its capabilities will remain irreducibly human, and has priced that decision as an investment rather than a cost. When the machines do the choosing, that decision is the whole strategy.
Research cited
Accenture, Me, my brand and AI: The new world of consumer engagement.
HubSpot, State of Marketing 2026.
VML, Future Shopper.
Ogilvy, The Human Premium.
Dentsu trend reporting.
Ashley Faus, Human-Centered Marketing: How to Connect with Audiences in the Age of AI (2025).
