What Makes an Engineering Manager Valuable When AI Can Do the Work?

I recently did an exercise that I would recommend to any engineering manager: I mapped out every activity I do in a typical week and asked, for each one, "Could AI do this?"

The answer was uncomfortable. About 50-55% of my work — ticket triage, sprint planning mechanics, status updates, release management, monitoring dashboards — is automatable. Not theoretically. Practically. Some of it I have already automated.

Then I did the same exercise for my manager. His number was closer to 35%.

The difference is not experience or seniority. It is where we each spend our time.

The automation gradient

Engineering management has a spectrum of activities, from purely operational to purely human. The operational end — moving tickets, pulling metrics, writing status reports, triaging bugs — is where AI excels. It is pattern-based, data-driven, and repeatable.

The human end — coaching someone through a career crisis, deciding which project to kill when two teams need the same engineer, convincing a stakeholder that their priority is wrong — is where AI has nothing to offer. These tasks require judgment under ambiguity, organizational trust, and the ability to read a room.

Most engineering managers spend the majority of their time on the operational end. That is not a criticism — the operational work is necessary and someone has to do it. But it is the work that is disappearing fastest.

What cannot be automated

After mapping my activities, I identified the work that no AI tool can touch:

Judgment under ambiguity. When two priorities conflict and there is no clear data to decide, someone has to make the call. AI can surface the data, frame the options, and even recommend — but the decision carries organizational weight that requires a human behind it.

Feedback calibration. Knowing that a direct report needs to hear "you are not being autonomous enough" is one thing. Knowing when to say it, how to frame it so it lands, and what to follow up with — that is calibration. It requires understanding the person, not just the performance data.

Organizational trust. People follow leaders they trust. Trust is built through consistency, vulnerability, and showing up when things go wrong. An AI cannot be the person who has your back in a talent review or who takes ownership during an incident.

Strategic framing. Deciding that your team should become "an automation-first engineering org" is not a data conclusion. It is a vision that synthesizes market signals, organizational dynamics, team capability, and timing. AI can inform it, but the synthesis is human.

The uncomfortable math

If 50% of your role is automatable, you have two options:

Option A: Fill the freed time with more operational work. Become the most efficient ticket processor in the company. AI eventually catches up. Your role shrinks.

Option B: Fill it with the non-automatable work — priority defense, people development, strategic decisions, cross-team influence. Your role grows into something AI cannot touch.

The engineering managers who will thrive are not the ones who resist AI. They are the ones who automate their operational work fastest — because that is what creates the space to do the work that actually matters.

A practical framework

Here is what I am doing about it:

  1. Audit your time. For one week, log every activity and categorize it: operational vs. judgment. Be honest about where you spend most of your hours.

  2. Automate the bottom. Pick the most repetitive operational task and automate it this week. Not next quarter. This week. A single script that saves 15 minutes per day changes your week.

  3. Invest in the top. For every hour you free up, spend it on the non-automatable work. Have the hard conversation. Write the priority rationale. Challenge the technical proposal. This is the muscle that atrophies when you are buried in operations.

  4. Track the ratio. I track autonomous decisions vs. escalations. The goal is 4:1. Every escalation should include my recommendation and reasoning — not just "what should I do?" This forces the judgment muscle even when it would be easier to defer.

The bottom line

The question is not "will AI replace engineering managers?" It will not — at least not the ones who do the human work. The question is whether you are spending your time on the work that matters, or on the work that is about to disappear.

Automate the 50%. Invest in the other 50%. That is the job now.