AI for Managers: The Four Things You Should Never Delegate

AI can do a lot of your job now. Which raises an interesting question if you’re thinking seriously about AI for managers: if AI can handle so much of the work, what’s actually left for you to do?

The honest answer is: more than you’d think, and it’s the part of the job that was never really about the work in the first place.

The Job Moves Upstream, Not Away

The instinct right now is to treat AI as a speed problem. Things that used to take a few hours take a few minutes. Things that used to take a team can sometimes get handled by one person with the right tools. Liz Fosslien and Mollie West Duffy describe exactly this shift in a Harvard Business Review piece on the AI productivity boom: execution gets faster, and managers are left struggling to keep pace with everything that speed makes possible.

That’s the part most people miss. Faster execution doesn’t shrink the manager’s job — it multiplies the number of things your team could be doing. More ideas. More projects. More requests. Somebody still has to decide which of those things actually matter, and that decision moves upstream, onto you.

Where AI for Managers Goes Wrong

The common approach right now is to treat every part of the job as a task to be optimized. If AI can summarize the meeting, organize the notes, and draft the first version, the logic goes, why not let it also handle the performance review, the one-on-one, the congratulatory Slack message? It’s more efficient. It’s less awkward. And it feels like the same category of work as clearing your inbox.

It isn’t. Some parts of management aren’t valuable because they’re difficult or time-consuming — they’re valuable because they’re human. They’re the moments where someone figures out whether they trust you, whether you understand them, whether their work actually matters to somebody. Hand those off and you’re not becoming a more efficient manager. You’re slowly removing yourself from the job of being a good one. If your team discovers that every “great job” they’ve received came from a chatbot, you haven’t scaled recognition — you’ve hollowed it out.

So here are the four things worth protecting on purpose.

Deciding What Matters

Jira can track the tasks. Asana can track the tasks. AI can now track the tasks, create the tasks, assign the tasks, and politely nag everyone about the tasks. None of that is your job anymore, and it probably shouldn’t have been.
What AI can’t do is decide what’s worth doing in the first place. That’s the judgment call — defining the core problem, clarifying what change you’re actually trying to create, deciding how success gets measured and who owns what. Use AI to inform that decision. Feed it context, ask it to surface options you haven’t considered, let it stress-test your reasoning. Just don’t let it make the call. Direction is leadership; task-routing never was.

Judging Your People

This is the one to be most careful with. AI can genuinely help you prepare for a performance conversation — pulling together your one-on-one notes, project results, peer feedback, and goals from six months ago faster than you ever could. Let it organize the evidence and remind you of patterns you might have missed.

But the evaluation itself — who has high potential, who deserves a raise, who needs improvement, who gets promoted — has to stay with you. Those aren’t numbers on a spreadsheet. They shape someone’s career, their compensation, their reputation, sometimes whether they still have a job. If you’re going to make a call like that about another human being, you need to be able to look them in the eye and explain it. AI can help you prepare for that conversation. It can’t have it for you.

Having the Conversation

There’s nothing wrong with running your thoughts through AI to get them a little more coherent before a hard conversation. The problem starts when AI becomes the thing your people are actually talking to, and you become the middleman relaying it.
A large part of leadership is simply paying attention to people — setting up the one-on-ones, asking how someone likes to work, what’s working, what isn’t, where they want their career to go. Companies love surveys. Managers love dashboards. It’s tempting to turn complicated human beings into percentages. But sometimes the most sophisticated management tool available is a real conversation, and no dashboard replaces it.

Creating Meaning

This one’s easy to hand over by accident, because AI is genuinely good at sounding encouraging. It can draft the congratulatory message. It can summarize the glowing customer testimonial. It can write a perfectly acceptable “Great job, team — really proud of what we accomplished.”

None of that is meaning. Meaning is knowing your contribution counts. Impact is knowing who’s counting on you. Put the two together and you get prosocial purpose — and helping people feel that connection is a core part of the job, not an administrative task sitting next to it. AI can help you find and organize the stories that carry that message. It can’t be the one who noticed.

The Real Line

None of this is an argument for using less AI. Let it summarize the meeting you didn’t need to be in. Let it organize the notes, draft the first version, find the patterns in the spreadsheet you’ve been staring at for three hours. That’s a legitimate use case, maybe one of the best available.

The line isn’t how much AI you use — it’s what you refuse to hand it. The more AI absorbs the administrative weight of management, the less excuse there is for skipping the human parts: deciding where the team is going, making the judgment calls about people, having the actual conversations. Those were never the annoying tasks sitting alongside the real job. They are the job. Getting the administrative half off your plate doesn’t shrink your role — it clears the space to finally do the other half well.