It's one person who pasted a spec into ChatGPT to see what came back, or used it to draft a method statement. That's curiosity, and it's healthy. But it's buying one cordless drill and calling it a fit-out.
A strategy is the boring part. Which job AI does first. Who owns it. What it has to prove before you trust it with the next one. What you’ll stop doing once it works. “We tried ChatGPT” answers none of those.
Individual experiments can be useful. A company-wide approach also needs agreed priorities, permitted uses, ownership and a way to assess what is working.
Dabbling tells you AI is interesting. A strategy tells you where it pays. Mistaking one for the other is a quietly expensive way to stand still.