Building Judgement as a People Manager

A few weeks ago I wrote about building judgement in People Operations now that AI handles the admin work where HR teams used to learn it. Several managers asked me what the same question looks like from their seat. The guidance changes in a few important ways.

What's different for managers

In People Operations, judgement mostly works across the organization: patterns, policy, and structure. A manager's judgement works at ground level. It's about a handful of specific people, every week:
‍

• Deciding whether a missed deadline is a one-off, a workload problem, or a sign that someone has checked out.
• Choosing who gets the stretch project and who needs a quieter quarter.
• Picking a rating for someone who had a strong year and one very visible miss.
• Knowing when to step in on a struggling project and when to let the person work it out.

Managers are also building their own judgement and their team's at the same time. Many were promoted from individual contributor roles in the last few years and are still learning the job while they do it.

Where AI is changing the manager's job

AI tools now sit in the middle of manager work. They draft performance reviews from a year of notes and goals, summarize 1:1s, suggest wording for feedback, and flag drops in sentiment or rising flight risk.
‍

Each of these saves real time. Each one also removes a moment where the manager used to have to pay close attention. Writing a review from scratch made you remember the whole year and decide what mattered. Taking your own 1:1 notes made you notice what someone avoided saying. When a tool does that first pass, it's easy to accept its version of events and move on.
‍

In the People Operations post, the main risk was that junior HR staff would miss the repetitive work that used to train them. For managers the risk is more immediate. Your own judgement can weaken even though you use it every day, because the tool keeps handing you a reasonable answer before you've formed one.

What carries over and what doesn't

Some of the original advice applies to managers with a different shape. Treating AI output as a draft to question still holds. So does giving people ownership of real decisions. The decision log still works, but it's mostly for you, not a teaching tool for a team.
‍

One piece doesn't carry over. People Ops staff need more exposure to the business. Managers already sit in the business. Their gap usually runs the other way: they see one team and rarely see how similar situations play out across the company. That's where HR comes in.

Building it on purpose as a manager

1. Decide before you read the draft. Before you generate a review draft or open an AI summary, write your own view in a sentence or two: the rating you'd give, the main strength, the one thing to work on. Then compare. Where the draft disagrees with you, find out why. Sometimes it caught something you forgot. Sometimes it's weighing the wrong things.
2. Keep a decision log for your own calls. When you make a call about a person, such as a promotion case, a performance plan, or giving someone a stretch role, write down what you expect to happen. Check back in 60 or 90 days. Seeing where you were wrong, in your own handwriting, is how you calibrate.
3. Treat AI flags as a reason to talk. A flight-risk score or a drop in survey sentiment tells you where to look. Finding out what is going on still takes a conversation with the person, and what you learn there builds your judgement for next time.
4. Ask for their call before giving yours. When someone brings you a problem, ask what they would do and why before you answer. AI makes it easy for team members to get an answer without asking anyone. Part of your job is making sure they're forming their own opinions and testing them against yours.
5. Use HR and peer managers to check your thinking. In People Ops, case reviews happen inside the team. Managers rarely get that. Bring a real situation to your HR partner or a peer manager before you act, especially on ratings, performance plans, and promotions. Calibration sessions do this formally. A 20-minute conversation does it informally.

Back to the design question

The original post argued that if we designed roles today, we'd design them around decisions instead of tasks. That applies to managers more than anyone. Take away the review drafting, note-taking, and status reporting, and what's left of the manager job is a set of calls about people: who grows, who's struggling, and what's fair. AI can give you better information for each of those. You still have to make the call, explain it to the person, and be accountable for how it turns out.