I learned that people in my organisation were using AI to help write performance and development documents. My reaction split cleanly down the middle: thrilled, and worried, in equal measure.
Thrilled, because this is exactly the maturity we’d been building toward. Nobody mandated it. People found a painful piece of knowledge work, the blank-page agony of self-assessments and development plans, and applied the tools intelligently. The documents I saw were often better than what came before: more structured, more articulate, less rushed at deadline.
Worried, because of what these particular documents are for. A performance conversation is one of the few moments we ask people to think carefully about themselves, what they did, what it meant, where they’re going. The writing isn’t the deliverable; the reflection is. The document is just the residue of the thinking. Hand the writing to AI and you might hand over the thinking with it, producing an articulate account of a year that its author never actually examined. Polished emptiness, in the exact place where authenticity is the whole point. And the person on the receiving end may respond with AI-assisted feedback, and we edge toward two machines exchanging pleasantries while two humans watch.
I chose not to make a rule about it. This felt like a moment for a conversation rather than a policy, so that’s what we did, openly. The distinction we landed on serves us well beyond performance reviews: use AI to sharpen your thinking, not to substitute for it. Think first, in rough honest notes, then let AI help you say it well. Never let it tell you what you did this year.
This won’t be the last time a use case is impressive and corrosive at once, and the boundary won’t usually survive being written as policy. Judgement has to carry it.
Where in your organisation is the writing actually the thinking? Those are the places to watch, and the conversations to have before the tools quietly have them for you.