"AI agents just means ChatGPT." It's a much wider, more personal category than that.
Most people picture one thing when they hear "AI agent": a chat window they open and type into, the way they'd open a search tab. That's one shape among many, and probably not the shape that ends up mattering most for buying and booking.
5 min read
Where the assumption comes from
A retailer planning its "AI strategy" for the year builds a single, specific picture in its head: someone opens ChatGPT, asks a question, gets a recommendation. Every decision gets made against that one scenario. It's not wrong, exactly, it's just a fraction of what "an agent" covers, and probably not the fraction doing the most buying on someone's behalf.
It's an understandable habit. For a couple of decades, "how do people find us online" meant one thing, a search engine, and businesses built their whole mental model around it. AI agents are getting slotted into the same mental shape: one dominant place people go, ask something, and act on the answer. That's a real, current habit. It's also just one distribution pattern among several already forming.
What the wider category includes
Some agents are standing relationships rather than one-off queries: an assistant someone's used for months, that already knows their shoe size, their dietary restrictions, their usual budget, and gets asked directly rather than searched fresh each time. Some are specialised, built to do one job, book a table, find a flight, rather than answer anything. Some aren't chat interfaces at all, they're embedded inside another product entirely, acting without a person typing anything into a window in the moment.
Why this matters for where effort goes
A business optimising only for "how does this look inside a chat window" is optimising for the most visible, most talked-about case. It's rarely the one that ends up buying the most. A standing agent that already knows a person's preferences has no chat window to persuade. It needs the same underlying facts every other kind of agent needs, reachable regardless of what shape the agent asking for them happens to take. It's the same mistake as designing for a page or a chat widget instead of the fact underneath: building for the interface everyone currently pictures, rather than for whatever shape actually shows up asking.
Where this shows up for a business
A business's facts get connected once through Selfe and stay reachable by whatever shape of agent asks for them, a chat window, a standing personal assistant, or something embedded inside another product entirely, rather than needing to be built for one assumed interface.
Do we need a different strategy for every type of agent?
No, the fix is the same either way: get the underlying facts right and connected. That works regardless of which shape of agent checks them.
Which type of agent should we prioritise?
Not one. The point of connecting facts directly, rather than building a chat-specific answer, is that it doesn't require picking.
A lot of what you've heard about agentic commerce isn't quite right.
Every myth in this hub comes from the same mistake: judging the future by how well it fits today's tools, a page, a widget, a chat window, a markup file, instead of asking what would make sense if you were starting clean. The actual destination underneath all of it is simpler than any of the myths: AI agents talking directly to data, not reading pages built for people.
"Bigger platforms will sort this out for us." Why we don't think that's coming, or coming for everyone.
The search era consolidated around a handful of gatekeepers, so it's a reasonable guess that agentic commerce will too, and that waiting for whichever one wins is the safe move. Genuinely autonomous, personal AI agents don't obviously converge that way, and waiting assumes a consolidation that might just not happen this time.
"Just add a chatbot" was never a strategy. Here's the argument for why.
A chat widget is a human-interface metaphor, a text box someone types into, bolted onto a business that hasn't changed underneath. It answers the question "how does a person talk to us" when the real question is whether the facts behind the widget are the ones an AI agent can act on.