Human intelligence is the edge AI can't buy

On 24 June we hosted the second OnlyBrands Live — sixty-odd brand-side marketers in a room in Manchester, four panellists from brands you'd recognise, and one question that sounded innocent and turned out to be the whole game. What happens to brands now everyone's pouring AI on everything? Notes from a long evening, the research that's catching up with what the panel said in the room, and why human intelligence — yes, that's the term we landed on — is becoming the only edge AI can't buy.

The room and the question

It started with statistics. Adweek's widely-cited analysis that 65% of the tasks performed by marketing professionals are eventually replaceable by AI — a finding that's been compressed in industry conversation into the punchier (and somewhat less accurate) line that 65% of marketing jobs may not survive AI in their current form. A figure from Writer's 2026 enterprise AI survey that 43% of marketing employees who already use AI at work believe their company would replace them with an AI agent tomorrow if it could. Roughly four in ten companies in the most recent surveys expecting to have replaced jobs with AI by the end of 2026, up from around one in five only twelve months earlier.

Then the conversation turned. Because once you've shown the numbers, the more interesting question isn't should we use AI. It's what's the work only humans can do, and how do we protect it before it gets quietly automated away?

The panel was deliberately mixed. Dr Amna Khan, consumer behaviour academic at Manchester Metropolitan University Business School and a trust researcher. Michelle Younger, Marketing Director at Aimia Foods, leading brands including Horlicks and Maltesers. Nick Moss, fifteen years at Astonish Cleaning Products and the literal face of the brand on its social channels. Charlotte Jackson, head of marketing at Lakeland Leather, deep in premium-fashion territory. Chaired by Guy Keeble of Reach plc. Sixty-odd brand-side marketers in the room. One overwhelming consensus by the end: human creativity is still the brand's only real competitive edge, and the brands ignoring that are already paying for it.

The AI slop problem (and what it really revealed)

The cautionary tale opens with the names you'd expect. Valentino's AI-led art campaign that produced what one observer in the room described as resembling a B-movie sleeve. Gucci's pastiche that read like a fever dream of Grand Theft Auto, Ghostbusters and Dynasty. Colgate publishing AI copy that hadn't even been proofread before going live. Each generated industry backlash measured in weeks of bad press and, in two cases, departures from the marketing team.

But the deeper pattern isn't the individual disasters. It's the convergence. Across categories, AI-generated visuals are starting to look the same — same lighting, same composition, same off-key human anatomy, same colour palettes. Brands that once invested in distinctive visual identity are sliding into a sea of beige. Byron Sharp's work at the Ehrenberg-Bass Institute on distinctive brand assets has been clear for years: brands grow by building and protecting mental shortcuts that make them instantly recognisable. AI-generated content, by its statistical nature, regresses to the visual mean. It's optimised for plausibility, not distinctiveness. We've written about the same convergence problem from the content side in stop making content to fill feeds, and made the broader case that the classic content marketing playbook is breaking down in what is content marketing? (and why most of it is dying).

The counter-movement was the more interesting story on stage. Fashion brands quietly making “we have hands” campaigns — celebrating the one thing AI still gets wrong, as a flag for craft. High-end labels leaning into harsh flash realism precisely because no AI was generating that kind of imagery yet. Dove's Campaign for Real Beauty, which marked its twentieth anniversary in April 2024 with a public pledge never to use AI to represent real women in its advertising. Aerie's headline-grabbing partnership with Pamela Anderson in March 2026, an extension of an October 2025 “100% Aerie Real” no-AI pledge that has run alongside a +23% Q4 comparable-sales lift. Content creators being asked to film the behind-the-scenes alongside the campaign — not as a bonus deliverable, but as proof of life.

And the most quoted line from Anthony's opening talk: 63% of consumers believe brands have a duty to disclose when they're using AI in marketing, according to research from consumer-panel platform Cint. Which sets up the trap nobody on the agency side had quite spotted.

The research nobody saw coming

Here's where Dr Khan's contribution to the panel started to bite. She walked the room through fresh research from Carney, Riveros and Tully (2026), an advance article published 7 May 2026 in the Journal of Consumer Research titled Made With AI: Consumer Engagement with Social Media Containing AI Disclosures. Eight preregistered experiments, plus TikTok content analysis, looking at what happens to consumer engagement when content is disclosed as AI-generated.

The headline finding wasn't what most marketers in the room expected. Consumers don't disengage from disclosed-AI content because they have a moral aversion to AI. They don't disengage because they think AI content is low quality. They disengage because disclosed AI signals a lack of effort.

Effort, in the research, is read as care. Care is read as relationship. And relationship — what social-psychology researchers call the parasocial connection — is what now does most of the work in keeping audiences attached to brands and creators. The Carney paper places parasocial-connection reduction as the central mechanism by which disclosed-AI content damages engagement, with the effort heuristic operating upstream of it. The authors also note that disclosures which themselves signal greater effort can mitigate the engagement drop. Disrupt that signal and audiences read it as the brand stepping back from the relationship. Engagement falls. Purchase intent falls. The follow rate falls.

That research lands inside a larger reshaping of how brand trust is built, which Dr Khan unpacked from her own PhD work on trust dimensions. Trust has historically had two dimensions: organisational trust (the brand's reputation, competence, expectation delivery) and personal trust (the emotional, individual relationship a person has with the brand or the people who represent it). Legacy brands have spent fifty years building the first kind, hoping the second would follow. Social media has flipped the order. Today, people often build personal trust with a founder, an influencer or an employee-creator first, and then trust the organisation behind them. The research she cited puts the personal dimension at roughly twice the impact of the organisational one.

Then she gave us Catherine from M&S as the canonical example. A product-development colleague whose face has become more familiar to UK consumers than most of M&S's paid creative — to the point where she now drives waiting-list demand for individual SKUs in a way M&S's brand voice on its own no longer can. Ben Francis at Gymshark. Paige at P. Louise. Skims, the brand built on Kim Kardashian as the trust vessel. Even Nike has had to bring an influencer-founder-equivalent into the marketing mix.

Put the Carney research and the parasocial trust shift together and the implication for AI is uncomfortable: the brands using AI most aggressively in their visible, audience-facing content are simultaneously eroding the relationship that brings audiences back. Optimising the wrong number.

What AI is actually good for

None of this means AI doesn't work. It works brilliantly in specific places. The panel were unanimous on where.

Octopus Energy now uses AI to handle roughly a third of customer email queries, and CEO Greg Jackson has publicly reported 80% customer satisfaction on the AI-handled threads versus 65% on the human-handled ones. The AI is faster, more patient, and — interestingly — more inclined to say yes. Holland & Barrett has deployed AI-powered communications platforms across its store estate to free shop-floor staff from routine product-lookup overhead, so the saved attention can go into deeper, more empathetic conversations with customers. AI-driven ad-variant testing has produced documented step-changes in performance: one widely-reported apparel-retail case study (FULLBEAUTY) recorded 45% higher return on ad spend, 22% higher conversion rate and 36% higher click-through using Meta's Advantage+ Shopping AI. And the Kalshi prediction-market spot that ran during Game 3 of the 2026 NBA Finals — made by AI filmmaker PJ Accetturo for around two thousand dollars in cloud compute, generating hundreds of clips across 48 hours to land 30 seconds of broadcast-grade creative.

What all those examples share: AI was doing the work the customer never sees, or AI was orchestrated by someone with the taste to know what to ask it for. Charlotte Jackson at Lakeland Leather described her own deployment as “invisible AI” — size recommenders, returns triage, search-bar question routing, customer service triage. Every interaction is AI-assisted; none of it reads as AI to the customer.

The principle the panel kept returning to: efficiency in the back, humans in the front. AI for the grind; people for the work that touches the audience directly. It's the same logic that drives our case for running organic and paid as one system — different tools, different jobs, one connected operating model.

The taste-maker problem

Anthony Diver, founder of fractional-leadership consultancy NAITIV and co-host of the evening, told a story that stayed with everyone in the room. A Manchester business with 150 staff, running a rigorous one-week AI bootcamp for every new starter. Chefs trained to use AI for allergen lookups. Universal adoption. Then he asked the founder: any pushback?

One person. The copywriter. He'd quit, leaving — the founder said with a laugh — his dusty thesaurus on the desk.

The laughs in the room died fast. Because letting the copywriter go without replacing the judgment that copywriter carried doesn't make the business more efficient. It makes it more average. The copywriter wasn't paid for grammar. They were paid for taste — the ability to know whether a sentence would land with a customer, which is the same skill a brand needs to know whether an AI output is good or just plausible. The same logic applies to creators: brief them like the taste-makers they are or you lose the thing you paid for. We covered that argument in most creator briefs kill the thing they paid for.

Dr Khan extended the same argument into cultural territory. Ask ChatGPT about Korean skincare, she said, and you'll get a location in South Korea. The model has no cultural fluency. It has training data. A brand manager's job is to be the cultural fluency — to know that “Korean skincare” is a category, a movement, a set of rituals, a price-tier expectation and a community before it's a location. The same applies to every category, every audience, every regional nuance, every generational difference. AI cannot read culture; it can only summarise the average of what's been written about it.

This connects to what behavioural economists like Rory Sutherland have called the Procrustean problem of optimisation — the temptation to cut whatever doesn't fit the measurable shape. Taste, cultural fluency, brand judgment: these things are real, expensive and difficult to measure. They get cut first when AI promises to make everything cheaper. And the cost shows up months later, as brand equity quietly degrades. We've written before about why customers want personality, not a script; the same argument now extends to AI-generated content at scale.

The orchestration argument follows directly. AI is a tool — sometimes a brilliant one. But the value isn't in the tool; it's in the person commissioning it. The Kalshi NBA ad cost two thousand dollars because the person making it knew exactly what to ask the AI for. Without that taste, AI delivers the same NBA ad anyone else could have produced, in eight hours rather than forty-eight, indistinguishable from a thousand other algorithmic outputs.

The hedonic edge

Some categories will always belong to humans because the experience itself is human. Michelle Younger made the point about Horlicks in two sentences: the number one driver of her brand is taste, and you cannot sample taste through a screen. The product proves itself in a sensory exchange that AI is structurally excluded from. The same is true for Maltesers, for fragrance, for fashion's texture, for hospitality, for restaurants, for any product where the moment of consumption is the moment of brand-building.

Charlotte's point about leather followed the same logic. Lakeland Leather sells products you need to touch. Imagery has to reproduce the texture as accurately as cameras and lighting will allow, because the customer will zoom in and zoom in and zoom in before they pay several hundred pounds for a handbag. AI-generated leather looks like AI-generated leather. The texture is hallucinated, not photographed.

The wider framing: every brand with a hedonic dimension — anything that engages the senses, anything that depends on physical presence, anything where the experience is the product — has an edge AI cannot erode. Brand commerce, in our framing, is the operating model that takes this dimension seriously rather than treating it as legacy overhead.

Founder-led trust and the power shift

The most important shift Dr Khan flagged on stage was about distribution. The traditional brand world ran on top-down distribution: brands paid for media, media reached audiences. Social media collapsed that hierarchy by giving every brand, every employee and every customer their own distribution network. The power moved.

The winners since the shift are founder-led brands that put a human face at the front and let it accumulate parasocial trust over time. Ben Francis at Gymshark. Paige at P. Louise. Kim Kardashian at Skims. Catherine at M&S. Nick Moss at Astonish — the example in the room — who built a brand-personality presence on Instagram over the better part of a decade by being personally identifiable rather than corporately polished.

This is also why the live shopping market has moved as fast as it has. Research that Dr Khan referenced suggests that Gen Z viewers stay on live shopping streams because they want to feel seen — by the host, by the chat, by the community. Trust rises with that feeling, and purchase intent follows. The Shopify or Amazon equivalent — browse anonymously, click to buy — cannot deliver the same feeling.

We've made the longer case for the community side of this in you can't manage a community. The point on stage was simpler: the shift from organisational trust to personal trust is permanent. AI doesn't reverse it; it accelerates the brands that already understand it and erodes the ones that don't.

What human intelligence means for brand commerce

This is where the evening landed for us specifically. Brand commerce — IF.'s founding idea, which we've defined formally in what is brand commerce? — is the practice of building brand and driving commercial performance as one connected discipline. The framing has never depended on AI being absent. It's about the operating model in which human intelligence is applied at commercial scale.

Read against the OnlyBrands evening, brand commerce splits the work into three categories that map directly onto the panel's instincts.

Efficiency work — the back-of-house, customer-invisible, optimisation-heavy operations where AI genuinely speeds up the business without affecting brand. Search bar routing, returns triage, ad-variant testing, customer service first-line, internal data analysis. AI belongs here, fully and unapologetically.

Execution work — the production, distribution and amplification of brand assets where AI can speed up parts of the workflow but a human taste-maker still commissions, approves and polishes. Creative direction, copy editing, image selection, video editing, social calendar construction. AI assists; humans decide.

Brand-defining work — the strategy, the cultural read, the founder presence, the live conversation with audiences, the moments of taste that distinguish a brand from a hundred similar ones. AI doesn't belong here at all. This is where competitive advantage actually lives.

The brands getting AI wrong aren't using AI at all. They're using AI in the third category — letting it write the copy, generate the campaign, run the founder's voice — and discovering, with the Carney research now to confirm it, that their audiences read the lack of effort and step back.

A practical playbook

Five principles, drawn from what the panel actually said and what the research now backs up.

Use AI where the customer doesn't see it. The Octopus, Holland & Barrett and Lakeland Leather pattern. Optimise the operations; protect the visible brand.

If you must use AI in customer-facing content, lead with the human craft. Disclosure is now legally and morally expected. The Carney research suggests that adding visible human craft — the “we have hands” gesture, the behind-the-scenes content, the founder showing up — restores the parasocial signal that pure AI content kills. The principle echoes our case for social content that actually sells: hooks, native format, real humans, every time.

Don't sack the taste-makers. The copywriter, the senior creative director, the cultural lead, the strategist — these are the people whose judgment makes AI outputs useful. Lose them and you lose the ability to tell good AI from average AI, which is the same thing as losing the ability to tell good brand from average brand.

Found and front your brand. Personal trust is now twice as impactful as organisational trust. Whichever face represents your brand publicly — founder, employee-creator, brand ambassador, regular human spokesperson — invest in their visibility and consistency. The compounding effect over a decade is enormous.

Protect the hedonic experience. Taste, touch, smell, sound, the felt quality of physical product or live service. These remain unrepeatable by AI, and they are the categories where brand commerce delivers its strongest returns. Don't let optimisation pressure cut what AI was never going to do anyway.

The bottom line

Anthony closed his opening talk with a phrase that became the unofficial slogan of the evening: elevate, don't replace. Use AI to take the grind off your team's plates. Use the time that frees up to do the brand-defining work AI fundamentally cannot do. Hire taste-makers and protect them. Front your humans and let them build the parasocial trust that's now doing the work organisational trust used to do.

That's the brand commerce read on human intelligence. The brands that win the next decade won't be the ones with the most aggressive AI adoption curve. They'll be the ones whose people are loud enough, distinctive enough and human enough to be cited by name — by their customers, by their competitors, and increasingly by the AI engines themselves.

If you'd like to think about what that operating model looks like for your brand, talk to us. Or come to the next OnlyBrands Live — third edition is in planning.

Sources and further reading

Carney, S., Riveros, I., and Tully, S. M. (2026), Made With AI: Consumer Engagement with Social Media Containing AI Disclosures, Journal of Consumer Research, published 7 May 2026 (DOI 10.1093/jcr/ucag013). academic.oup.com/jcr. Referenced by Dr Amna Khan at OnlyBrands Live.

Adweek (2026), “65% of Marketing Jobs May Not Survive AI” — the 65% figure draws on Anthropic's analysis of marketing task automatability.

Writer (2026), “The AI leadership gap: Even marketers who use AI fear they'll be replaced”, drawing on Writer's 2026 AI Adoption in the Enterprise Survey.

Cint (2026), “63% of consumers think brands have a duty to disclose when they use AI in marketing”.

Dove (April 2024), Keep Beauty Real pledge, twentieth anniversary of the Campaign for Real Beauty.

Aerie x Pamela Anderson “100% Aerie Real” (March 2026), Marketing Dive: how Aerie is pushing back against AI content with Pamela Anderson.

Greg Jackson / Octopus Energy on AI customer service, via The Wrap: AI satisfies customers better than humans.

Kalshi NBA Finals AI ad case study via DesignRush: Kalshi NBA Finals AI ad.

AI ad-creative benchmark data (FULLBEAUTY case study) via Digital Applied: AI ad creative benchmark 2026.

Byron Sharp and the Ehrenberg-Bass Institute, How Brands Grow (2010, 2015) — on distinctive brand assets and mental availability.

Les Binet and Peter Field for the IPA, The Long and the Short of It (2013) — on the 60/40 brand-versus-activation budget split.

Rory Sutherland, Alchemy (2019) — on the Procrustean risk of over-optimisation.

Edelman Trust Barometer, annual reports — background context on the long-term decline in institutional trust.

IF.'s related thinking: what is brand commerce?, what is content marketing?, what is social commerce?, how to make social content that actually sells, stop making content to fill feeds, most creator briefs kill the thing they paid for, customers want personality, not a script, you can't manage a community, the performance plateau, organic builds, paid scales.

FAQ

What is human intelligence in brand marketing?

The judgment, taste, cultural fluency and emotional connection that human marketers, creators and founders bring to brand-building work — the things AI can support but cannot replicate. The term reframes AI not as a replacement for human marketers but as a tool that lets them concentrate on the brand-defining work only they can do.

Should brands disclose when they use AI in marketing?

Yes, and they're increasingly legally obligated to. But the Carney, Riveros and Tully (2026) research in the Journal of Consumer Research shows that disclosed AI content reduces engagement because it signals lack of effort, which damages the parasocial trust relationship. Brands should treat disclosure as compulsory and use AI selectively in customer-facing content.

Where should brands use AI safely?

In invisible, back-of-house operations: search and recommendation systems, returns and customer service triage, internal data analysis, ad-variant testing, optimisation. Keep AI out of brand-defining work like strategy, creative direction and founder voice — that's where personal trust is built, and where AI use most damages it.

What is brand commerce's view on AI?

Brand commerce treats AI as a tool that frees human intelligence to concentrate on the work that builds brand and drives sales together — strategy, creative direction, founder voice, cultural fluency. AI handles the grind; humans handle the brand-defining work. The result: efficiency without erosion of the personal trust that now drives consumer behaviour.