AI Guides › Career & Work

Using AI To Draft Your Own Exit Interview Answers, Honestly

By Nigel Guy · 2 min read

The default instinct in an exit interview is diplomatic smoothing — "great experience, just time for a change" — regardless of what actually happened. It feels safer in the moment. It also means the organisation never hears anything it could act on, and you leave a conversation that could have been useful having said almost nothing real.

The rule: use AI to separate what's true from what's safe to say out loud — draft the honest version first, then decide deliberately what to soften, rather than smoothing by default before you've even named the real answer.

The mechanism: the Honest Draft

  1. Write the unfiltered version first, for yourself only. What actually made you leave, without softening — this draft is never sent anywhere, it exists so you know what you actually think before deciding what to say.
  2. Sort each point into say plainly, say carefully, or leave out. Not everything true needs to be said, and not everything said needs to be blunt — the sorting is the actual work, not an afterthought.
  3. Rewrite the "say carefully" points as specific and constructive, not vague and safe. "Communication could be better" helps no one. "Decisions about X were made without the team that had to execute them" is something an organisation can actually act on.
  4. Decide what you leave out and why, deliberately — protecting a relationship, avoiding a legal grey area, or simply not wanting to relitigate something already settled are all legitimate reasons, but they should be a choice, not a default.
  5. Rehearse saying the carefully-worded version out loud before the actual conversation, the same way you'd rehearse any other hard conversation — see "Using AI To Practise Saying No At Work" for the same underlying method.

What to skip

Skip using the unfiltered first draft as your actual answers — it exists to clarify your own thinking, not to be read aloud. Skip inventing specific criticism you don't actually believe just to seem candid; the goal is honest, not performatively blunt.

Guardrails

All 751 AI guides · JulieMango plans from £17/mo