When companies first explore AI, the economics seem obvious. A general-purpose AI tool costs a few hundred euros a month. A custom AI assistant costs tens of thousands to build. The comparison isn't even close.
Except it's not really a comparison. Because a generic tool and a custom assistant don't do the same thing — and the gap between them has a cost that doesn't show up in the licensing fee.
The adaptation tax
Generic AI tools are built for the average use case. If your workflow is average, they perform reasonably well out of the box.
But most workflows aren't average. They have specific terminology, specific processes, specific edge cases. To close that gap, teams spend time configuring, prompting, and workarounding. Someone — usually a smart person who has other things to do — becomes the de facto AI wrangler.
This time has a cost. It's just invisible because it shows up in salary, not in a software invoice.
The adoption gap
Generic tools also suffer from an adoption problem that's easy to underestimate.
When an assistant behaves inconsistently — sometimes useful, sometimes off-base, never quite predictable — users stop trusting it. And when they stop trusting it, they stop using it. Not because they decided to stop, but because they quietly developed a habit of not bothering.
Low adoption is the most common outcome for generic AI tools in professional environments. The team didn't reject it. They just didn't build a habit around it.
The total cost of generic
When you account for adaptation time, ongoing maintenance, inconsistent outputs, and partial adoption, the real cost of a generic tool is often higher than the licensing fee suggests.
The question isn't "what does the tool cost?" It's "what does the problem cost if the tool doesn't really solve it?" — and then "what would it cost to actually solve it?"
That's not always an argument for building custom. But it's always worth doing the full calculation before assuming the cheaper-looking option is actually cheaper.