The chatbot is the easy part

Every SMB I talk to opens with the same ask. 'We want a chatbot.' Sometimes it is a chatbot on the website. Sometimes it is a chatbot in Slack. Sometimes it is a chatbot that sits on top of a knowledge base nobody has cleaned up in three years. The shape changes. The instinct does not.

Here is the thing nobody tells you when you are shopping for AI. The chatbot is the easy part. It is the thinnest layer of the whole system. You can spin one up in an afternoon with the right API key and a prompt that fits in a Slack message. The hard part is everything underneath it.

The hard part is what data it can see. What it is allowed to write back. Which workflow it interrupts versus which one it replaces. Who owns the prompt when the founder gets bored of editing it. What happens when an answer is wrong and a customer is sitting on the other end. None of that is solved by picking a better model.

I have a working theory. Call it the surface vs substrate problem. The surface is the chatbot, the form, the widget, the thing your team sees and gets excited about in a demo. The substrate is the workflow change underneath. Surface is cheap and fast. Substrate is slow and political and the only place real value lives.

The teams I have seen win with AI did not start with a chatbot. They started with one workflow they were tired of. Patient intake. Sales order generation. Vendor security questionnaires. They picked the workflow, mapped it end to end, and only then asked which surface made sense. Sometimes the answer was a chatbot. More often it was a form, a script, an email-watcher, or a tool that nobody outside the team ever sees.

If you are an SMB founder looking at AI right now, the question to ask is not 'where can we add a chatbot.' It is 'which workflow do my people complain about every week.' Start there. The surface picks itself once the substrate is honest.