For most of my career, handing over work has been emotionally difficult. I always worried that they would screw it up. What fewer of us have reckoned with is that we now feel exactly the same reluctance when it comes to handing work to an AI agent.

The interesting thing is that the emotional barrier is often larger than the technical one. In the interview study I have been conducting on how professionals adopt agentic AI, the feedback is quite consistent: people use these agents in minimal, cautious ways and as assistants. There seems to be a threat to professional identity. If the agent can do the thing I have built my professional identity around, then who am I? So we keep the agent on a short leash, supervise every step, and never quite let it run. And by doing so, we capture only a fraction of the value on offer.

I think the parallel is a lesson about good leadership: The best leaders I have experienced set a clear, valuable outcome and then get out of the way. That does not mean blind trust, of course. It means learning where the agent can be trusted and where it cannot, and expanding that boundary deliberately over time. And then, much like we feel about our children, being genuinely proud when the agent does the job better than we ever could. So, as a reflection for today: what work are you clinging to that you should be letting go of? And is it because the agent cannot do it, or because you are not yet ready to admit that it can?