When a mid-sized company decides to “do something with AI,” the request that lands on our desk is almost always the same: a chatbot. On the website, in the app, answering customer questions. It’s the first thing everyone pictures.
It’s also, most of the time, the wrong place to start.
Why everyone asks for the chatbot
The chatbot is the AI you can see. It looks like the demos. It’s the thing your competitor just put on their homepage. So it becomes the default ask — not because someone traced a costly problem back to it, but because it’s the most legible form of “we have AI now.”
That’s the tell. The chatbot is usually a solution looking for a problem, and the problem it’s attached to is “we want to look modern.” That’s a marketing goal wearing an engineering costume.
What a chatbot actually costs
A chatbot is not a one-time build. It’s a thing you now own.
It needs a knowledge base that stays current — every outdated answer is a support ticket and a trust hit. It hallucinates confidently, which means you’re now managing the risk of your brand telling a customer something false in writing. And it resets expectations: once there’s a bot, customers expect it to actually work, and a mediocre one is worse than none.
For most companies, the honest first-year cost of a chatbot isn’t the build. It’s the maintenance and the reputational downside nobody budgeted for.
The automations nobody asks for
Here’s what would have moved the needle more, for less: the work no customer ever sees.
The three hours a day someone spends sorting inbound emails into the right queue. The quote that takes two days because it’s assembled by hand from four systems. The invoices classified manually, one by one. None of these are exciting. All of them have a number attached — a salary, a delay, an error rate — and that number is what makes them worth automating.
Invisible automations don’t demo well. They just quietly pay for themselves, which is the entire point.
When a chatbot is the right call
Sometimes it genuinely is. If you have real support volume, a high share of repetitive questions, and a knowledge base that’s already good and maintained — a well-scoped assistant can take real load off a team. The difference is that you arrived there from the problem (support is drowning) rather than from the solution (everyone has a bot now).
So before you brief anyone on a chatbot, ask one question: what costly, nameable problem is it solving? If the answer is “it’ll look good,” you’ve found a better use of the budget somewhere quieter. Here’s how we figure out where AI actually pays off.