Will AI replace managers, or just make management harder?

A large share of management is administrative: scheduling, status reports, summarising, chasing, translating one team's update for another. All of that is now automatable, and the honest response is that most of it should be.

What is left when you remove it is not a smaller job. It is a more concentrated one, made entirely of the parts people find hardest.

What AI genuinely takes off a manager's plate

Status collation, meeting notes, first drafts of reviews and plans, rewriting the same update for four audiences, and the enormous amount of scheduling work that fills a calendar without advancing anything.

This is real relief and it is not trivial. For a lot of managers it is most of the week, and getting it back is the difference between managing and administering.

What is left is the hard part, and it is now the whole job

Deciding what the team should work on and defending that decision upward. Telling somebody their work is not good enough in a way that makes it better. Noticing that a quiet person has checked out. Carrying accountability for an outcome you did not personally produce. Choosing between two people who both want the same opportunity.

None of these are information problems, which is why none of them are helped much by a system that supplies information. A manager whose week is now made entirely of these is doing a harder job than a manager whose week was half administration, even though it is a shorter list.

The administrative half of management lifting away, leaving the concentrated remainder: deciding what the team should do, telling somebody their work is not good enough, and carrying accountability for an outcome somebody else produced. — A shorter list, and a harder week.

The new problem: evaluating work you cannot see the making of

Managers used to assess capability partly through output. It was imperfect and it correlated. That correlation is gone, and most management practice has not noticed.

This creates a specific failure: the highest-output person on the team may be the most fluent tool user rather than the most capable colleague, and a manager rewarding output will promote them. It takes a long time to discover, and the discovery usually happens during a crisis.

The response is to assess reasoning rather than artefacts. Ask what was considered and rejected. Ask what would change the conclusion. Ask what the weakest part of the argument is. Then check, later, whether what somebody predicted turned out to be true.

A manager reading two pieces of work of equal polish, one from deep understanding and one from fluent tool use, with nothing in either artefact to tell them apart. — Rewarding output now promotes whoever is best at producing it.

Managing a team where everyone has a machine

Two things change practically, and both need saying out loud rather than assuming.

The first is that the team's output no longer tells you the team's health. You need direct signals: what people are stuck on, what they are learning, whether anyone is producing work they do not understand.

The second is that development stops being automatic. The routine work that used to build judgement is being done by the tools, so growth has to be arranged deliberately or it does not happen at all. That is now a core part of the job rather than a nice thing to do when there is time.

Common questions

Will AI replace managers?

It replaces the administrative half of management: status collation, scheduling, summarising, first drafts. What remains is deciding what matters, handling people, and being accountable for outcomes, none of which are information problems. The job gets smaller in hours and harder in substance.

How should managers evaluate performance when AI writes the work?

By reasoning rather than artefacts. Ask what was considered and rejected, what would change the conclusion, and where the argument is weakest. Then check predictions against outcomes over time. Output volume has stopped being evidence of capability.

What new skills do managers need in the AI era?

Enough AI literacy to know what the tools are reliably bad at, the ability to assess reasoning in conversation rather than through documents, and deliberate development of their people, because the routine work that used to grow juniors is now automated.

Is management getting easier or harder with AI?

Easier in hours, harder in substance. The relief is real, and what is left is the concentrated difficulty: judgement, difficult conversations, accountability, and developing people whose usual learning path has been automated away.