Field note2026-09-15T15:35:35.873Z

"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.

5 min read

The Selfe team
Agentic commerce infrastructure
A friendlier front door doesn't fix what's behind it. The widget was always the easy, visible half of the job.

What a chat widget fixes

A cottage-letting agency adds a chat widget to its homepage, trained on its own FAQ pages, and reports the project as done. Visitors can now type a question and get an answer instead of scrolling to find it. That's a genuine, if modest, improvement for a person on the site. It has almost nothing to do with an agent booking on someone else's behalf.

A chat widget shortens the distance between a person's question and an answer already sitting somewhere on the site. That's real, and worth having. But it's a convenience layer over the same underlying facts the site already contained. If the cancellation policy was vague before the widget, it's still vague, now with a friendlier front door.

What it doesn't touch

An agent booking on someone's behalf isn't a visitor typing a question into a widget. It's a system checking specific facts, directly, often without any conversational interface involved at all. A chat widget answers the question "can a person ask us something in plain language." It says nothing about whether the underlying fact that answer depends on is accurate, current, or reachable by something other than the widget itself.

Why "just add a chatbot" keeps getting suggested anyway

It's an easy thing to sell and an easy thing to point at: a business can demo it, screenshot it, show a board member. Fixing the underlying facts, making sure the cancellation policy, the dog policy, the actual room count are accurate and connected everywhere they're checked, is less demoable and more of the real work. The widget is the visible part of a project that skipped the invisible, harder part underneath it.

Where this shows up for a business

Selfe fixes the layer underneath the widget: a cottage agency's real facts, connected and current, reachable by an agent directly rather than only through a chat window built for a person typing a question.

So a chat widget is a waste of time?

Not for the people using it. It's a separate project from agent-readiness. A useful one, just not the same one.

What should a business do instead?

Check whether the facts a chat widget would answer from are accurate and connected everywhere they're needed, not just inside the widget's own FAQ.