In our previous articles on the 6 GenAI spend traps, we've covered model routing, hybrid pipelines, and pilot overload.

This article focuses on Spend Trap #4: output quality driving rework.

This is, in some ways, the most insidious trap in the series. The previous three eventually show up in your AI budget. This one doesn't. It shows up in people's time, and people are significantly more expensive than tokens (unless you've horrendously misconfigured your agent teams).

The pattern is familiar: GenAI produces output that looks right: it reads well, is structured nicely, and then someone with domain expertise spends 45 minutes correcting the details, rewriting the conclusions, checking the citations, removing the confidently stated nonsense, and wondering whether it would have been quicker to just write it themselves.

When that "someone" is a senior consultant, a lawyer, or a compliance officer, that 45 minutes is not cheap. Particularly when it's a lawyer.