Have an AI prep your one-on-one, and bring questions, not a verdict

sweetmelon team · Management

The one-on-one is at three. You cannot remember what happened to the blocker they raised last time, or whether you ever did the thing you said you would chase for them. The traces are all there, spread across the channels you were already in. Have them pulled onto one page before the meeting: what they raised, which of those items went quiet afterwards, what you promised, which of their requests are still sitting unanswered somewhere. Facts and a timeline, nothing more. Do not let it tell you who seems disengaged or whose tone has changed lately. That kind of read is the easiest thing for a model to produce and the hardest for you to check, and walking in already holding a verdict about someone is how the conversation stops being useful. Keep it to the channels you can already see, too. Shared threads, tickets, project updates. Reaching into private messages to prepare for a conversation with the person who wrote them is a different act, and it will be understood as one. What the page gives you is a set of things to ask about, none of them settled. "Did that dependency ever clear?" is a question. "I can see you were stuck for two weeks" is a conclusion you reached without the person in the room. You stop opening with how are things going, and problems surface a few weeks earlier than they used to, because the question was specific enough to answer.