When the Buyer Is a Bot, Your Brand Magic Never Loads
The shopping is being handed to agents. They cannot see your hero video. They can only read your data, and they are not impressed by either.
There is a quiet shift in who you are actually selling to, and most brand teams are still optimizing for the wrong audience.
HBR's March piece, Preparing Your Brand for Agentic AI, lays out three ways customers now shop through AI: they talk to a brand's own agent, they send a third-party agent they have trained over time, or they let their AI deal with other AI on their behalf. A Kearney survey cited in the piece found 60% of US shoppers expect to make purchases through agentic AI within the next 12 months.
Sit with the second and third modes for a second. In those, the thing evaluating your brand is not a person. It is a model. No nostalgia, no weakness for a beautiful film, no patience for your homepage. It reads structured data and makes a call.
The agent cannot see your magic
Almost every dollar a brand spends goes into the part an agent ignores. The cinematic spot. The emotional tagline. The art-directed hero image. That work moves humans, and it will keep mattering wherever a human is still in the loop. But when the customer's own agent runs the first pass, your brand magic never loads. What loads is your product feed, your specs, your price, your availability, your reviews, your policies. The machine-readable facts. HBR calls the asset you need for this a "brand code," a structured knowledge base both people and agents can act on. Most brands do not have one. They have a brand book in a PDF and a data layer held together with tape.
This is a data problem in a marketing costume
Here is what that work tells me about the agentic shift: the brand that wins the human funnel can still lose the agent funnel completely, and not because its product is worse.
Because its data is a mess.
"How?", you ask.
Four product feeds that disagree. A price that is right on the site and wrong in the structured markup. Inventory the agent cannot verify, so it quietly routes to a competitor it can. The agent does not punish you for being unconvincing. It punishes you for being illegible.
Forrester is watching the same shift from the infrastructure side. Its read on the next-generation CDP is that the customer data platform is turning from a dashboard humans query into infrastructure agents operate, reading profiles, deciding, and acting in seconds. If your data is not clean enough for your own agents to act on, it is nowhere near clean enough for someone else's.
Winning in a new digital twin world
The auto world already has a live example, and HBR uses it: Capital One's Auto Navigator concierge can check dealer inventory, schedule a test drive, estimate a trade-in, and answer financing questions. That works because the data behind it is structured enough to act on.
The brands that agents select will be the ones that treat their structured presence, the feeds and specs and availability and policies and reviews and consent flags, as a first-class product instead of the unglamorous afterthought it is in most orgs today.
None of this means brand stops mattering. If anything, it means brand matters more than ever. Agents may pick whatever is most structured, but humans are still emotional and irrational consumers.
The brand that's able to marry differentiated, emotional messaging to strategic digital activations will intercept customers while they're still in the consideration or decisioning phase before an agent decides for the customer. It's the difference between being a market wunderkind or a market has-been.
It means brand now has two audiences with opposite tastes. One wants to feel something. The other wants to verify something, fast, and will route around you without a flicker of guilt if it cannot. The job is to stay legible to the machine without going forgettable to the human, and almost nobody is even trying yet, which is the whole opportunity.
The customer used to be the one you had to convince. Now, we must brave the eight layers of AI miasma before we even get to the customer. Never mind the fact that often the customer is asking unrelated questions of an AI model that eventually gets them to a product that we never would have considered being relevant search terms to optimize or purchase against.
Customers treat AI like the verification backstop that their friends and family used to be. Anyone who's been in advertising long enough will tell you that word of mouth from a trusted source is one tough obstacle to overcome.
So, here we are, once again proving that a technological revolution didn't make our jobs easier as intended, instead doubling the work. Either we build for both, or get verified out of the running before a human ever lays eyes on our brands.
Anthony
← More writing · If this was useful, or wrong in an interesting way, tell me: hireme@crmjedi.com.
Get the next one by email
New essays on customer data, measurement, and what actually ships. No spam, no cadence theater.