Margin Note

The Brand in the Prompt

The Brand in the Prompt. Architectural exploded view on aged tracing paper. Bottom layer shows technical spec sheets and barcodes in precise gray, a routing mechanism sits above them, and floating at the top a luminous brand signet in amber. A small agent sensor reaches past the spec layers directly toward the amber signet, a line leads across to a prompt input field on the side.

68% of consumers have already used ChatGPT to shop. McKinsey estimates the agentic commerce market at 2.5 to 4.2 trillion euros by 2030. The accompanying thesis from VML's Tomorrow's Commerce Research: when agents do the buying, only data matters. Specs, prices, availability, machine-readable and complete.

"An AI agent evaluating your product doesn't care about your heritage, your campaign, or your brand ambassador," writes my colleague Paul Appelhof.

Stay up to date

Get notified when I publish something new, and unsubscribe at any time.

For batteries, that's true. For printer paper. For anything where the purchase is emotionally unloaded. There, the cleanest dataset wins.

For everything else, in my view, it's fundamentally wrong.

A brand isn't an ornament on a product. It's a compressed value profile: trust, status, belonging, quality promise, condensed into a single signal. Which is exactly what an agent needs when it has to decide on behalf of its human. A template it can match against the profile of the person it works for.

Red Bull objectively tastes borderline. But "Red Bull" encodes something an agent can read: energy, risk, performance. If the agent knows its human watches extreme sports and gets up at five, Red Bull matches. No dataset in the world delivers that without the brand behind it.

The second mechanism is simpler: people write their favorite brands into the prompt. "Order me running shoes" is an open search. "Order me Hokas" is a decision, made in the head of the human, not in the algorithm. Byron Sharp showed that 95% of potential buyers aren't in the market at any given moment. And when the remaining 5% become ready to buy, the decision in favor of the brand with the highest mental availability has effectively already been made. Agents amplify the effect: strong brands don't just land in the prompt, they land in memory. And what's in memory doesn't get compared again.

The agentic commerce literature frames the agent as a wall between brand and consumer. The opposite is more plausible: the agent is a concierge. It knows its human's wishes and uses brands as matching templates. It brings the brand closer to the person, not further away, because it's the only intermediary that understands both sides.

Machine-readability is the entrance ticket. The brand is the reason the agent recommends you.

Whether incumbents and challengers will have to play strategically different games in this world is its own topic. For now, just this: if you're already in the head, you'll land in the prompt. If you're not, you have to find another way into the agent.

Ask yourself: If an agent shops for your best customer tomorrow, is your brand in the prompt? Or only in the database?