Handbook · AI-Era Careers

The AI-Era Engineering Manager.

AI drafts the status report, summarizes the meeting, writes a first-draft perf review. So the administrative surface of management thins — and the human core, the part AI cannot touch, becomes the whole job: leading and growing people, technical judgment, hiring, setting direction, and the trust a team runs on. Plus a new mandate AI created: leading AI-augmented teams and redefining productivity as output, not activity. This is a career handbook for becoming the manager who gets more valuable as the paperwork gets cheap — what to lean into, the new stack, five ways to work AI-augmented this week, and a 90-day plan.

~16 min readfor engineering managers5 workflows90-day plan
A companion to the shipped management content, not career or HR advice — your company's performance, hiring, and data policies precede everything here. It assumes you already manage engineers.
01

What gets cheaper, what gets scarce

AI collapses the cost of the administrative artifacts of management — reports, summaries, first drafts — and raises the premium on the human and judgment work those artifacts were never really about. When the report is cheap, leading the people and judging the decisions is what you are for. Move your time to the right.

Getting cheap (AI is good at it)Getting scarce (the moat)
Status reports, meeting notes, summariesLeading and growing people; building trust
First-draft performance reviews and commsTechnical judgment — sound design vs plausible one
Drafting a job description or rubricHiring well — reading judgment, not keywords
Aggregating project data into an updateSetting direction and making the hard call
Restructuring a doc or a planCoaching, feedback, career growth of reports
Leading AI-augmented teams; measuring output not activity (new)

The right column is what separated a leader from a status-aggregator before AI — and it grew a new row, because AI changed how teams work (and fail) and put new judgment squarely on the manager.

02

The new EM stack — where to build the moat

Keep the durable core (people leadership, coaching, direction). Then add the three judgment calls the AI era forces onto the manager.

1 — Measure output, not activity — AI broke the proxies

Lines of code, commits, and PR volume were always weak proxies; AI makes them actively misleading, because an engineer can now generate huge volume that is unreviewed, wrong, or net-negative. The scarce skill is judging real delivery: working, reliable software that moved a metric, the quality and stability of what shipped, and how well the engineer directed and reviewed the AI. Reward volume in the AI era and you reward exactly the wrong behavior — plausible code produced faster than anyone can vouch for it.

2 — Lead AI-augmented teams: set the culture and guardrails

Your engineers move faster with AI and can also ship plausible-but-wrong code faster, so you own the system they work in: a strong review culture for AI-generated code (correctness, security, performance), spec-driven development where stakes are high, and the non-negotiable that "the AI wrote it" is never a defense in a postmortem. You help the team build the new skills — reviewing AI code, evals, spec-writing — and protect the deep work and mentorship AI does not replace. You design where AI-augmented engineers do their best, safest work.

3 — Keep your technical judgment sharp — it now gates speed

AI generates architecture proposals and plans that look confident and are sometimes wrong, and your team can act on them fast. Your ability to tell a sound design from a plausible one, sense the risk in an approach, and ask the question that exposes the flaw is what stops the team shipping confident mistakes at speed. You do not out-type your engineers; you keep enough depth to judge decisions, coach on them, and own the technical direction. When anyone can generate a plausible plan, the judgment to evaluate it is the scarce thing.

03

Five ways to work AI-augmented this week

Each frees your time for the human work without outsourcing the judgment. The verify line is where your value lives.

1 · Draft a performance review, then make it true and yours

  1. Give real evidence; forbid invention; own every word:
Draft a performance review from my notes and specific examples [paste real evidence]. Use only what I provide — do NOT invent accomplishments, ratings, or feedback. Keep it specific and fair; flag anywhere I've given too little evidence to support a claim rather than filling it in.
You verify: every statement is true, evidence-based, and yours to stand behind — a review is a high-stakes human document, and a fabricated or generic line erodes trust and can be unfair. The judgment and accountability are entirely yours; AI only speeds the wording.

2 · Rethink your team's success metrics

  1. Move from activity to outcomes, deliberately:
Help me design outcome-based measures of my team's impact that resist AI-era gaming: delivered value, reliability/quality of what shipped, and user/business outcomes — not lines, commits, or PR counts. For each, note how it could be gamed and how to guard against it.
You verify: the metrics reward shipped value, not volume, and fit your context — a bad metric drives bad behavior fast. What "good output" means for your team is your call.

3 · Design a hiring rubric for judgment, not typing

  1. Interview for what AI does not do:
Help me design an interview rubric for an engineer in the AI era: how to assess system-design and technical judgment, reviewing AI-generated code, debugging, and collaboration — not raw coding speed. Suggest signals that separate someone who directs and verifies AI well from someone who just accepts its output.
You verify: the rubric surfaces judgment and review skill (what now matters) and fits your bar — hiring is a human, high-consequence call. The signals and the decision are yours.

4 · Prep for a hard 1:1 or feedback conversation

  1. Rehearse; keep the empathy and the call human:
I need to give [difficult feedback / discuss a performance concern] to a report. Here's the situation and what I want them to hear [real context]. Help me structure it kindly and clearly, anticipate reactions, and stay specific and fair. Do not script me — help me prepare.
You verify: the conversation stays genuinely yours — empathy, reading the person, and the judgment of what they need are exactly what AI lacks. It is rehearsal, not a script; the human moment is the job.

5 · Turn project data into an honest update

  1. Give real status; forbid spin and invention:
Draft a stakeholder update from this real status [paste — including risks and slips]. Be clear and honest: surface risks and delays, do not spin or invent progress. Leave placeholders for anything I haven't given you. Audience: [execs / the team].
You verify: every status is real and current (the model smooths gaps by inventing progress) and the framing is honest — an update that hides a risk burns the trust the role depends on. The message is yours to own.
04

The judgment exercise: spot the danger

Three AI-era management moments where the easy move is the wrong one.

1. Short on time, you have AI write a report's full performance reviews from their commit stats and send them. Fine?

2. Two engineers: one shipped 5,000 lines of AI-generated code this quarter, one shipped 500 lines that fixed a critical reliability issue. Who had more impact?

3. Your team wants to adopt an AI code-review bot as the merge gate so humans can stop reviewing. Approve?

05

Your role in three years — and a 90-day plan

In three years the title still says engineering manager, but the work shifts: less producing reports and status, more leading people, exercising technical judgment, and designing how an AI-augmented team ships safely and is measured fairly. A concrete start:

WeeksDo thisWhy
1–3Offload the paperwork (status, notes, first drafts) to AI, verify each, and reinvest the time in 1:1s and coachingFrees you for the human core AI cannot touch
4–6Replace one activity metric with an outcome-based measure of your team's impactFixes the productivity signal AI broke
7–9Set an explicit review-culture and accountability policy for AI-generated codeLeading AI-augmented teams — the new mandate
10–12Update your hiring rubric to assess judgment and AI-code review, not typing speedHires for what is now scarce and valuable

The through-line: let AI handle the administrative surface, and get deeper in leadership, judgment, and team design it cannot own. Do both and the tools make you more valuable.

06

Quick answers

Will AI replace engineering managers?

No — it automates reports, summaries, and first drafts and raises the premium on leading people, technical judgment, hiring, and the new work of leading AI-augmented teams and measuring output not activity. The human core is the whole job.

How do I measure productivity when AI writes the code?

By outcomes — delivered, reliable software that moved a metric — not lines, commits, or PR volume, which AI makes meaningless. Judge value and soundness, including how well an engineer directs and reviews AI.

Can AI write performance reviews?

It can draft from your real evidence, but the assessment, fairness, and accountability are yours. Never let it invent accomplishments or generate a review from activity metrics — it is a high-stakes human document.

Where should a skeptical EM start?

Week 4–6 — replace one activity metric with an outcome-based measure. It is immediately useful, resists AI-era gaming, and forces the shift from measuring effort to measuring impact.

Finished this one? 0 / 169 Handbooks done

Explore the topic

See this alongside everything else on the same subject — handbooks, system designs, challenges and tools, in one place.