21/04/2026
1. Avoid anthropormorphization, a.k.a Claude is not your colleague. Don't use personal pronouns to refer to AI - stick to "it". "It" reminds us that it is a tool that can put to a particular use. If you have a screwdriver, you would not expect it to be able to saw a log in half. What you have is a machine capable of making highly sophisticated stastical inferences. You should not expect it to be a machine that predicts the truth.
2. Do not present the output of AI without a framing. When sharing output of AI, include what your perspective is on the information. Is it, to your knowledge, accurate? How confident are you of that? Otherwise the other person might go away thinking that they have certified information, when actually it's a vague guess.
3. Signal that something was produced by an AI. If you write a JIRA ticket, sign it off as written by Claude under your instructions. When explaining an idea, start "Chatting with Claude, I came up with an idea to...". Developers, make sure that a commit is signed as co-authored by Claude.
4.