The one rule for working with AI in a team

While there are a lot of things one can learn about how to more effectively use AI with a team - there is really just one hard and fast rule:

That's it.

If there's a corollary to that rule, it's to mark everything that was wholly AI generated as AI generated. If you've really edited and made it your own, you don't need to do that (although it might be good to still call out AI's role).

Because of the inversion-of-effort in effect with how quickly LLMs can generate text, you need to make sure you aren't shifting the burden of working with AI to your peers.

If you want to have AI generate a 40 page document, that's fine, but you need to read and refine that document yourself and make sure it's right. If you aren't willing to read that 40 page document, then it sure doesn't seem right to make your colleagues read it either. If you send it to your boss - there will likely be incorrect information or conclusions in there. If you send it to subordinates, you'll likely create confusion and discussion as people try to figure out what certain vague, but correct-sounding sentences mean.

Because LLMs are good at writing text that seems right, a cursory once-over isn't enough. I've seen folks read a slide bullet from their own presentation and watched them become confused. They had asked an LLM to create a slide, had given it that cursory glance, but then when it came time to present the slide, the meaning was clearly not what they had intended. Don't be that person.

For more specifics - see AI shouldn't be writing for you.