HR has a new colleague to think about

This week has been mostly about AI and HR. Across several sessions, I kept coming back to what HR already knows about helping someone become useful in an organisation. Purpose, context, responsibility, feedback. There is a lot here to work with.

The onboarding and offboarding deck makes that connection concrete. Give an agent a role, appropriate access and a named human owner. Introduce it to the organisation's language and approved sources. Start with supervised work, look at what people have to correct, and decide how much responsibility it has earned.

I suggested that an onboarding process could start by asking whether you are onboarding a human or an AI. Then the paths can differ. HR has experience to bring to both. The question I want to keep in view is what success would actually look like, beyond getting the new colleague through the door.

Ask what we could learn from them

There is a part of human onboarding that bothers me. We spend so much time explaining our culture and our way of working. How often do we ask a new colleague what worked better where they came from?

That fresh perspective has a short window. People can eventually stop pointing out the things everyone else has grown used to.

I want to bring that question into working with AI too. Explain how you intend to approach a process, then ask for another way. Review where the agent lacked context, which sources were missing and what your instructions made unclear. The conversation can improve the organisation's way of working as well as the agent's next attempt.

Keep the learning when an agent leaves

Offboarding deserves attention from the beginning. When an agent is replaced, what happens to the instructions, corrections and useful methods you developed together? Who takes over its unfinished work? Who withdraws its access?

I argued this week for capturing that value and leaving a sensible recovery path when trying a replacement. A new tool should be able to benefit from what the previous collaboration taught you.

These questions are part of the territory I am exploring through AIHR. The reading path on onboarding an AI colleague is a place to continue. Try it with one real task in mind. Who owns the result, what would make it worth keeping, and what would you want to retain if it ended?