Judgment > Tools
Using AI isn’t proof of an advantage.
I would evaluate what becomes usable, how much correction it needs and who owns the decision—not just which model a company uses.
“We use ChatGPT” tells me very little about how a company works.
Imagine two sales teams with access to the same model. One produces a proposal quickly, but someone still has to rewrite the scope, check the price and chase approval. The other defines what an acceptable proposal includes, supplies the right information and sets a review step before anything goes out.
Same access. Very different work.
That is the comparison I care about: how long it takes before something is actually usable, how much correction it needs, and who owns the final decision.
For a first test, I’d take a small set of completed cases, remove sensitive information and compare the current process with an AI-assisted version. I’d track errors and review time, not just generation speed. OpenAI · Evaluation best practices.
Maybe AI improves the process. Maybe a simpler form does.
Either outcome is useful. What I wouldn’t count as progress is producing more drafts while the same person stays late fixing them.
Show me a piece of work that now gets done properly, with less effort. That tells me much more than the name of the tool.