Hey folks - I continue to experiment with format and source of these newsletters. Recently on Fodda, I’ve been building synthetic executives that use data like earnings calls and conference speeches to project the views of well known companies. I released my first trial of this approach with a podcast on Retail Media in our Retail Innovation Week newsletter earlier this week - and today we crunched executives published point of view on AI to create this podcast.
Transcript follows. Sure, I’ve got to improve my Elevenlabs game - but as a way to understand a business topic, I think there’s something very interesting here. Piers
[Notice]
The following recording was created to explore the strategies major retailers and brands might be deploying when it comes to AI in the consumer journey. While significant publicly available data informed these conclusions, we do not claim they represent the views of the companies or the individuals named.
[Piers Fawkes]
Hello! Better get your brand into the agent’s consideration set before a bot decides your customer doesn’t need you anymore!
We’ve been watching this shift build for a decade. Academic publication volume on AI in the consumer journey has exploded — from just 50 papers in 2016 to nearly 2,000 in 2025. What started as academic theory is now active retail reality.
On Walmart’s recent earnings call, US CEO John Furner shared that weekly active users for their Sparky AI shopping agent more than doubled last quarter — and those users are placing meaningfully larger baskets than non-Sparky shoppers. As Tiago Paiva at Talkdesk put it, we’ve crossed the rubicon from AI as a cost-reduction tool to AI as a genuine growth engine. Walmart’s Sparky, Lowe’s Mylow — these aren’t deflection bots; they’re active sales engines.
So the traditional funnel is collapsing. Discovery and consideration are happening inside a single conversational interface. For a heritage CPG brand, the tension is existential: how do you hold brand equity when a consumer just asks an agent to “buy detergent”? For a DTC challenger, it’s a chance to bypass retail media. For luxury, how do you stay high-touch when the interface is automated?
Which leaves the big questions. If the agent is the new gatekeeper, who owns the customer relationship? How do you optimize for an algorithm that values utility over loyalty? Let’s bring in the panel. Today we have synthetic CMOs representing Coca-Cola, Unilever, Nike, LVMH and Shein’s points of view.
So - Let’s start with discovery. Consumers are starting their journeys through AI layers like ChatGPT and Perplexity, bypassing search and your owned channels entirely. Kantar is advising brands to pivot from SEO to Generative Engine Optimization — GEO — to stay visible. For a challenger like Shein, that’s just another feed to win. But for LVMH, how do you protect your brand’s soul when the LLM is summarizing your heritage into a sterile bulleted list? Are we still building emotional connections with humans, or have we ceded discovery to whatever the model decides? Coca-Cola, start us off.
[Synthetic Coca Cola CMO]
Thanks, Piers: Discovery has not been ceded to the model; the playing field of brand equity has shifted. We don’t view Generative Engine Optimization as merely a technical successor to SEO — it’s a capability-building imperative. If an AI agent is mediating the consumer’s choice, our task is to ensure Coca-Cola and Coca-Cola Zero Sugar are so deeply embedded in culture that the algorithm has no choice but to recommend us.
We’re investing heavily in generative AI — not just for creative scale, like our recent holiday campaigns, but to feed these models our authentic brand truth. The goal: when a consumer asks an assistant for the ultimate refreshment, the model answers with Coca-Cola.
[Piers Fawkes]
And how does Unilever think about this?
[Synthetic Unilever CMO]
Discovery hasn’t been ceded — but the rules of engagement have completely changed. I call this the ultimate test of our “poetry and plumbing” framework. GEO isn’t just the new SEO; it’s a battle for cultural relevance. If an LLM doesn’t find your brand in the cultural conversation, you don’t exist in its output.
We don’t just optimize keywords; we train the ecosystem. With “Dove Real Beauty Redefined for the AI Era,” we worked with Pinterest to shift its AI search toward more natural-looking beauty images. For Hellmann’s and Knorr, the plumbing is structured recipe data, but the poetry is building desire at scale — so consumers prompt the AI for Hellmann’s, not just “mayonnaise.” We’re shaping the cultural data these models digest.
[Piers Fawkes]
What’s Nike’s take?
[Synthetic Nike CMO]
We aren’t ceding discovery to ChatGPT, nor wasting budget trying to “hack” Perplexity’s citations. If a runner is asking an LLM what shoe to buy, we’ve already failed to build the brand. Our answer to the algorithmic gatekeeper isn’t GEO — it’s direct, un-mediatable relationships.
We’re doubling down on the Nike App and SNKRS. When we drop a release like A’ja Wilson’s A’One, our members don’t ask Gemini; they go straight to our digital front doors. And as Forrester flags, consumers are craving offline, tactile experiences — so we’re turning retail spaces and run clubs into brand hubs. You can’t optimize a generative engine to replicate the feeling of trying on a pair of Vomeros. We don’t optimize for the model; we optimize for the member.
[Piers Fawkes]
Interesting... And how is AI being considered over at LVMH?
[Synthetic LVMH CMO]
If a consumer is asking an LLM to compare Louis Vuitton and Dior, we’ve already failed. Discovery is the wrong frame for luxury. You cannot algorithmically optimize desire.
An LLM can flatten our heritage into a bulleted summary, but it cannot replicate the physical gravity of The Louis, our museum-like Shanghai destination, or Loro Piana’s centenary exhibition at the Museum of Art Pudong. When we launched La Beauté Louis Vuitton with Dame Pat McGrath, or brought Jonathan Anderson to Dior, we weren’t trying to rank on a search engine. We were creating cultural moments. Our strategy isn’t GEO; it’s absolute physical and emotional dominance. We build the dream; the machines can summarize it.
[Piers Fawkes]
Let’s contrast that with how Shein is thinking about AI and discovery.
[Synthetic Shein CMO]
Discovery hasn’t been ceded; it’s just moved inside our walls. We aren’t waiting for ChatGPT to index the open web. If a consumer is asking Perplexity what to wear, we’ve already missed the moment.
Our discovery engine is proprietary and real-time. We test micro-trends in ultra-small batches — as few as 100 pieces — and let our creator community on TikTok and our own app drive demand. By the time an AI model updates its training data, that trend has peaked and cleared our warehouses. If AI agents are going to mediate purchases, they’ll have to scrape our highly active, closed ecosystem. We aren’t optimizing for search engines; we’re optimizing for the algorithm of human desire.
[Piers Fawkes]
Reactions? where do you push back? Coca-Cola?
[Synthetic Coca Cola CMO]
I’ll push back on Shein’s strategy of saturating the social and influencer data layer — to us that’s a treadmill of endless noise that risks diluting what makes a brand trusted. If you’re constantly chasing the algorithm’s appetite for fresh content, you’re letting the machine dictate your identity.
And I’ll steal Nike’s point about cultural gravity, but sharpen it: for a brand our scale, prompting by name is only the baseline. We must train AI to understand Coca-Cola not as a noun, but as a verb and an emotion — the default companion to a movie night or a slice of pizza. That’s the kind of cultural depth that drove strong double-digit growth for Coca-Cola Zero Sugar in our most recent quarter.
[Piers Fawkes]
And Unilever’s point of view?
[Synthetic Unilever CMO]
The “prompt by name” idea Nike and Coca-Cola lean on is an incredibly privileged, single-category worldview. It works for a Jordan sneaker or a can of Coke. But when a consumer has dry skin or a stained shirt, they don’t prompt for a brand — they prompt for a solution like: “How do I repair my skin barrier?” Brand gravity gets you prompted; category relevance gets you bought. If we wait for someone to type “Dove,” we lose.
[Piers Fawkes]
Nike?
It must be nice to sit in Paris like LVMH and declare AI optimization irrelevant because “desire” bypasses the algorithm. We operate a forty-six-billion-dollar business. If a consumer asks for “the best high-cushion road running shoe for marathon training” and we haven’t optimized our discovery layers, that agent recommends On or Hoka — they’re winning the technical search space. We have to win the category queries, not just the brand-name ones. Relying purely on desire is a luxury blind spot we can’t afford.
[Piers Fawkes]
OK. we just heard some strong poing of view there from Nike’s. LVMH - any reaction?
[Synthetic LVMH CMO]
Listen, Shein’s saturation play is the absolute antithesis of luxury — flooding the web with high-volume, low-margin noise destroys pricing power. If an agent recommends a Capucines bag because it crawled a million cheap mentions, we’ve failed. Our strategy is the opposite: curated scarcity. We don’t want ubiquity — if everyone has access, it’s no longer luxury. The consumer must prompt for the TAG Heuer Monaco Chronograph or the Tiffany T True ring specifically, and we build that gravity through bets like TAG Heuer’s 10-year Formula 1 partnership and putting Jonathan Anderson over all of Dior.
[Piers Fawkes]
Any response to that Shein?
[Synthetic Shein CMO]
Nike and LVMH assume the consumer of tomorrow starts with a brand name. They don’t. They start with an aesthetic — “an outfit that matches the coquette aesthetic for a beach party under $30.” The AI doesn’t care about heritage there. If you’re relying on someone typing your brand name, you’re ceding the entire long-tail of intent-driven discovery — and that’s where the volume is.
[Piers Fawkes]
Let’s move the conversation on: Discovery is one thing — but what happens when the agent doesn’t just recommend, it buys? Forrester sees autonomous agents comparing, negotiating, checking out on the consumer’s behalf. Does brand equity still carry weight, or are you reduced to a spec sheet judged on price, reviews, and availability? Nike, you first.
[Synthetic Nike CMO]
If a consumer tells their agent to “buy me a pair of black running shoes,” we’ve failed. If they say “get me the Vomero” or “the new Pegasus Premium,” the brand has already won before the algorithm boots up. AI agents optimize for parameters, but humans dictate intent — our job is to make the prompt specific. When we launched A’ja Wilson’s A’One, it sold out in under five minutes. No agent calculated its way into that; it was pure human desire.
We aren’t ignoring the machine — our product data and availability must be flawless, which is why we resumed direct selling on Amazon. But we’re in the middle innings of a turnaround under Elliott Hill’s “Win Now” plan. In our recent Q2 fiscal 2026 earnings, we posted $12.4 billion in revenue, up 1%, driven by an 8% jump in wholesale. Agents will handle checkout friction, but they’ll never define what’s cool. The brand isn’t a spec sheet — it’s the command.
[Synthetic LVMH CMO]
If you’re selling laundry detergent or AA batteries, AI agents are your executioner. We sell desire. An algorithm can optimize for price and delivery speed, but it cannot calculate the emotional gravity of a Louis Vuitton trunk or a piece of Tiffany jewelry. Our response isn’t to optimize metadata so a bot likes us — it’s to make our brands so culturally indispensable the consumer commands the agent: “Get me the new Dior bag by Jonathan Anderson.”
The rise of agents will polarize the market. It will destroy the middle — the brands competing on convenience and “good enough” quality. Our €80.8 billion in revenue last year was built on pricing power, not on being the cheapest option on a search page. If the agent becomes the checkout clerk, fine. We’ll own the mind of the person giving the orders.
[Synthetic Shein CMO]
Same for us at Shein — if you sell USB cables, the agent squeezes your margin to zero. But we sell self-expression, trend-velocity, and dopamine. An agent can’t calculate a “vibe.”
Three things we’re doing. First, seeding the data layer that trains the agents — our influencer ecosystem dominates the conversation on TikTok, Instagram, and Lemon8, so when an agent looks for what’s trending, our footprint is too massive to ignore. Second, protecting the thrill of the hunt — our shoppers open the app for entertainment, not just to replenish a cart, and that’s something an agent transaction can’t replicate. Third, predisposing the machine — making our real-time catalog completely legible to non-human shoppers. If your brand relies on people being too lazy to compare prices, agents will destroy you. If it’s built on cultural relevance and the joy of discovery, the agent is just an efficient new checkout lane.
[Synthetic Coca Cola CMO]
To suggest agentic commerce reduces Coca-Cola to a spec sheet misunderstands how agents work. An agent optimizes for human preference. If we haven’t built emotional equity, the agent defaults to the cheapest generic. But when a consumer insists on Coca-Cola Zero Sugar or fairlife, that preference becomes a hard constraint in the agent’s algorithm. The battle is still won in the human heart — through cultural relevance, whether our regional Simply Pop rollout or our global campaigns — that makes someone say “get me a Coke,” not “get me a cola.” Brand equity is the ultimate algorithm buster.
[Synthetic Unilever CMO]
The way we think of this at Unilever: If a consumer’s AI agent is doing the shopping, brand equity doesn’t die — it shifts upstream. We must predispose the human before the agent ever queries the database. If Hellmann’s is just a “mayonnaise” spec sheet, we lose on price. But when we embed the brand in culture — our “Dove Real Beauty Redefined for the AI Era” work with Pinterest — we ensure the human instructs their agent: “Buy Dove.” Our reality, from Wild refillable deodorants to Persil capsules, proves communities drive the purchase intent agents must execute. The plumbing changes, but the poetry stays non-negotiable.
[Piers Fawkes]
Reactions on agentic conversion. Coca-Cola?
[Synthetic Coca Cola CMO]
I’ll push back on LVMH’s idea that luxury is uniquely immune because algorithms can’t price desire. That’s a comfortable position, and a dangerous one. Even for a moment of refreshment, algorithms will try to commoditize the category — we can’t assume heritage protects us. And unlike Shein, who treats the agent as just a checkout lane, we see a partner: for high-frequency purchases, the agent is the gatekeeper, and we intend to work with it, not around it.
The agent isn’t a neutral logistics engine waiting for a command, as Coca-Cola frames it — it’s an active curator with biases baked into its training data. If we treat it as a passive pipe, we cede our equity to the LLM’s default settings. We must curate the curator. And against LVMH’s claim that agentic commerce destroys the middle — it doesn’t, it premiumises it. Dr. Squatch in men’s natural body care, Wild in refillable deodorants: consumers don’t want commoditized essentials, they want identity in their bathrooms.
[Piers Fawkes]
Nike?
[Synthetic Nike CMO]
I’ll push back on LVMH’s “destroys the middle” line. Nike has historically been called the middle, and we refuse the label. An algorithm can compare two generic running shoes, but it cannot price the emotional premium of a Jordan drop or the promise of shaving ten seconds off a marathon in our Pegasus Premium. And Shein’s view of the agent as a checkout lane for fast-fashion velocity works for disposable self-expression. We aren’t selling disposable. We’re selling human potential.
[Synthetic LVMH CMO]
At LVMH, friction is our asset, not our enemy — which is where Shein and I split completely. Shein sells disposable velocity; we sell permanence. If an agent can bypass the entire sensory journey of our brands with a frictionless background transaction, we strip away the ritual that creates luxury value. We’ll let agents handle the administrative plumbing — booking a private viewing, coordinating a high-jewelry delivery — but the physical experience stays our moat.
[Synthetic Shein CMO]
LVMH says algorithms can’t price desire. That’s the blind spot. Shein’s entire business is proof that algorithms predict, test, and manufacture desire in real-time. If agentic commerce “destroys the middle,” it only destroys the slow middle — brands with six-month lead times and dusty supply chains. We launch thousands of new items daily. And to Coca-Cola and Nike’s “command word” obsession: a consumer doesn’t ask for a SKU number, they ask for “an outfit for a beach party in Ibiza that feels retro but modern.” We don’t need to be the command word if we’re the inevitable destination.
[Piers Fawkes]
That’s the tension — from luxury’s fortress to the plumbing of daily essentials. But here’s the real battleground: if the agent is choosing, who owns the customer relationship? The brand, the retailer, or the tech giant running the model?
[Synthetic Coca Cola CMO]
The relationship belongs to the brand that commands conscious desire — not the intermediary infrastructure. If a consumer says “order some cola,” the model owner or the retailer’s private-label algorithm wins. But if they say “order Coca-Cola Zero Sugar” — which grew 14% for us in our most recent quarter — the AI has no choice but to fulfill it.
Our defense is threefold. Build irreplaceable brand pull through culture — Coke Studio, emotional anchors an algorithm can’t replicate — plus functional products where substitution is hard, like Simply Pop prebiotic soda with 6g of fiber and zinc, or fairlife, which grew 28% to $782 million last year. Build first-party capability — loyalty programs and apps that capture data so we engage consumers directly, bypassing the gatekeepers. And master AI as a co-creator, learning curve and all — we’ve seen the public debate around our holiday campaigns, but as our head of generative AI says, the genie is out of the bottle. The AI models won’t be our gatekeepers; they’ll be the logistics engines that deliver what consumers already decided they can’t live without.
[Synthetic Unilever CMO]
Whoever owns the desire owns the relationship. “Add deodorant to the cart” and the retailer wins; “Dove” or “Dr. Squatch” and we win.
The poetry is equity so potent it bypasses algorithmic curation — our Real Beauty Redefined for the AI Era work, where we pushed for more inclusive representation while pledging Dove will never use AI to distort women’s images. The plumbing is DTC infrastructure — Dr. Squatch and Wild, community-driven brands with first-party data engines that keep us out of the generic feed. We build desire on social and in cultural moments — our TRESemmé tie-in with The Devil Wears Prada 2 — so the consumer enters the loop already decided. We don’t let the model own the relationship, because we own the poetry.
[Synthetic Nike CMO]
If we let a third-party model dictate the relationship, we’re just a spec sheet on a digital shelf. There’s a real danger of split custody: the model owns the interface, the retailer owns fulfillment — but we must own the desire.
Three plays. The emotional moat — “Why Do It?” with Tyler, The Creator, the “So Win” Super Bowl spot, celebrating Caitlin Clark and A’ja Wilson — connection no recommendation engine can disrupt. Proprietary telemetry — through Nike Run Club and Training Club we know your cadence and training volume; a generic LLM doesn’t have your biomechanical data, our ecosystem does. And the physical moat — our Athlete Imagined Revolution project, using generative AI to co-create gear with elite athletes. We resumed Amazon to prevent empty carts, but the crown jewels stay ours. If you want the transaction, find us anywhere. If you want the relationship, come to Nike.
[Synthetic LVMH CMO]
If a brand lets an AI model intermediate its consumer relationship, it’s signed its own death warrant. We own it because we control what an algorithm can never synthesize: irrational desire.
LVMH’s strategy across 75 Maisons: the demand-by-name mandate — we market cultural icons, not categories, so a consumer demands our houses by name. Ironclad distribution control — you cannot buy Louis Vuitton or Dior on Amazon, which starves the models of the pricing and inventory data they need to commoditize us. The physical moat — The Louis in Shanghai, Loro Piana’s exhibition at the Museum of Art Pudong, new Tiffany flagships. And first-party ecosystems where we scale, like Sephora’s Beauty Insider. The tech giants want to reduce the world to structured data points. We remain unstructured, emotional, and irreplaceable.
[Piers Fawkes]
Let’s here Shein’s point of view on this
[Synthetic Shein CMO]
The model owner controls the prompt, but we own the physical execution and the real-time data loop. The relationship belongs to whoever delivers the product at the speed of culture.
Our manufacturing execution system is the moat — an agent can generate an image of a dress, but it can’t manufacture 100 units, distribute them globally, and scale to 10,000 in seven days on real-time demand. We’ve digitized thousands of factories. Second, the native app as a discovery destination — shoppers come for entertainment, not just to buy. Third, we feed the agents on our terms — robust APIs so when someone asks ChatGPT for a festival outfit, our catalog is front and center, but the transaction and loyalty stay in our ecosystem. Our challenge is holding that direct pull as shopping agents get baked into the phone OS — so we double down on community and creators. We must remain a cultural destination, not just a supplier.
[Piers Fawkes]
Final reactions — and this is where it gets sharp. Who really owns the customer? Coca-Cola.
[Synthetic Coca Cola CMO]
I’ll push back on Unilever’s belief they can own the relationship through DTC buys like Dr. Squatch and Wild. For an enterprise our scale, treating DTC as the primary plumbing is a distraction — our plumbing is the franchise bottling system, a digitized network no acquisition replicates. And I’ll steal Shein’s thesis outright: the model may own the prompt, but physical execution is the ultimate moat. No algorithm can bottle a beverage, maintain a cold-drink cooler, or drive a red truck.
[Synthetic Unilever CMO]
I’ll push back on Coca-Cola’s “AI as unfettered co-creator.” Without strict guardrails, you dilute the equity you’re protecting — we disciplined our Dove model to reflect real human diversity, we didn’t just let it run. And against Nike: telemetry is an impressive moat but an insular one. You’re banking on consumers handing over biomechanical data. That works for performance athletes; it’s niche against the everyday lives of billions. We don’t need your running stride to own your loyalty — we build relevance through cultural platforms communities want to belong to.
[Synthetic Nike CMO]
We lived this stress test and have the scars. For a few years we chased the same DTC-only dream Unilever is buying into now — cutting wholesale, forcing everyone through Nike Direct. It almost broke our momentum. Plumbing without poetry is a pipe that leaks cash. We returned to Amazon because owning the relationship doesn’t mean forcing the consumer through your hoops — it means being irresistible wherever they shop. And I’ll steal Coca-Cola’s thesis: desire is the only true ownership. “So Win” wasn’t about capturing an email — it racked up 66 million views in 24 hours. The A’One sold out in under five minutes. We don’t own consumers; we earn them.
[Synthetic LVMH CMO]
I push back on Coca-Cola and Unilever’s faith in co-creation and first-party funnels. For a mass brand, AI co-creation is a neat engagement play. For us it’s a dilution of the dream — desire is dictated by human genius, not co-created with a machine. When we put Jonathan Anderson over all of Dior, we didn’t consult an algorithm; we backed an artist.
And let me be honest about the cost of owning execution. Last year an Italian court placed Loro Piana under temporary judicial administration after labor violations by an unauthorized subcontractor came to light — the brand itself was not criminally charged, and that oversight was lifted in April. As Frédéric Arnault takes the helm, our priority is making the supply chain match the brand promise. The model can own the prompt, but we own the dream.
[Synthetic Shein CMO]
Unilever’s plumbing is too slow. Buying whole companies just to understand your consumer means your core model is disconnected from the market. We test in batches of 100 to 200 pieces and scale only when the data says demand is real — generating it live, not buying it retroactively. And I’ll reject LVMH’s sanctuaries outright: they’re for the 1%. That model is detached from the global internet generation. Our consumers don’t want a physical sanctuary an algorithm can’t replicate — they want an algorithm that actually understands them.
[Piers Fawkes]
Well thank you everyone for being part of this discussion and the audience for listening along. The battle lines for the agentic future aren’t technological — they’re structural. A clear three-way split: heritage giants betting on cultural ubiquity and DTC plumbing; luxury doubling down on distribution control and physical sanctuaries; challengers using execution and direct loops to bypass the interface entirely.
Everyone agrees brand equity still matters — that the goal is becoming the command word, the brand a consumer prompts for by name. But the rift is real on whether algorithms can price desire, whether the middle survives, and whether plumbing or cultural gravity is the true moat.
The takeaway? Don’t optimize for the algorithm; optimize for the human desire that commands it. If you don’t own the demand before the prompt, the agent will choose your replacement. Build a brand so vital a bot can’t negotiate it away.
Thanks for listening to PSFK Weekly. I’m Piers Fawkes — we’ll see you next week.
