Will AI replace engineering managers?
No, but it is taking a big bite out of the job and some of the headcount. AI already does much of the status reporting, coordination and documentation that used to fill an engineering manager's week. Large companies are using that slack to remove manager layers and widen spans of control. What's left is the part of the job that was always the hard part: hiring, developing people, handling conflict and owning trade-offs.
So the honest answer has two halves. The role survives. Fewer people will hold it, each with more reports and more AI output to answer for, and the version of the job that was mostly routing information between people is the one that disappears.
I've managed engineering managers for a long time, and in my experience the work divides into two piles of roughly equal size. One pile is information plumbing: who's blocked, what shipped, what's at risk, what the VP needs to hear by Friday. The other is judgment about people and trade-offs. AI is very good at the first pile. That's the mechanism behind everything in the data below.
What does the data on manager cuts actually show?
The measured evidence is about flattening, not replacement. The large-company moves are real and documented. The bigger economy-wide claims are still predictions.
| Source | What it says | Type |
|---|---|---|
| Google (Aug 2025) | 35% fewer managers with small teams than a year earlier; many stayed on as ICs | Reported internal figure |
| Amazon (Sept 2024) | Target to raise the ratio of individual contributors to managers by at least 15% by end of Q1 2025 | Stated goal |
| Gartner (Oct 2024) | Through 2026, 20% of organizations will use AI to flatten structure, eliminating more than half of current middle management positions | Prediction |
| Microsoft Work Trend Index (Apr 2025) | Introduces the "agent boss" and a human-agent ratio as a new org-design metric | Survey and framework |
| Gallup | Managers account for at least 70% of the variance in employee engagement across business units | Research estimate |
Google's number is the most specific. In August 2025, Brian Welle, Google's VP of people analytics and performance, told employees the company had 35% fewer managers with small teams than a year before, as NBC News reported. The cut targeted managers with very few direct reports, and many of them moved into individual contributor roles. Amazon's version came earlier: Andy Jassy's September 2024 memo set a goal to raise the ratio of individual contributors to managers by at least 15% by the end of Q1 2025, which Fortune covered as an anti-bureaucracy push.
Neither company said "AI replaced these managers." Both framed it as removing layers and bureaucracy. But the timing isn't a coincidence. Wider spans have always been possible in theory and painful in practice, because every extra report adds status meetings, one-on-ones and write-ups. When AI takes over a chunk of that overhead, the span that felt unmanageable at eight reports starts to look workable at twelve.
The forecasts go further. Gartner predicted in October 2024 that through 2026, one in five organizations would use AI to flatten their structure and eliminate more than half of their current middle management positions. Gartner also flagged the costs: remaining managers overloaded with direct reports, and junior staff losing the mentoring that managers provided. Treat it as a prediction about a minority of organizations. It's the headline people quote, and it is not a measurement.
Which parts of the engineering manager job does AI absorb?
The coordination and reporting work: anything where the manager's main job is to collect information from one place, compress it, and deliver it somewhere else.
| Task | What AI does now | What's left for the manager |
|---|---|---|
| Weekly status reports | Drafts from tickets, PRs, deploys and incidents | Deciding what the risk actually is and saying it plainly |
| Meeting notes and follow-ups | Transcribes, summarizes, extracts action items | Deciding which meetings should exist |
| Dependency and blocker chasing | Flags stale tickets and blocked PRs automatically | Resolving the conflict behind the blocker |
| Sprint and ticket hygiene | Writes tickets, groups work, suggests estimates | Saying no to work and protecting focus |
| Performance review prep | Gathers evidence: shipped work, reviews, incidents | The judgment, the calibration and the conversation |
| Onboarding Q&A | Answers "how does X work" from docs and code | Building the new hire's relationships and confidence |
The second-order effect matters more than any single task. A lot of middle management exists because information doesn't flow on its own. When status, risk and progress become queryable, the manager is no longer the only place to get them. A VP who can ask a tool "what's at risk this sprint?" needs fewer layers between them and the work. That's the structural reason flattening and AI adoption show up together.
What doesn't AI replace?
The parts where a person has to make a call about other people, and own it. These were always the core of the job, and the reporting work just crowded them out.
Hiring and team composition
AI can screen and summarize. It can't decide whether a candidate will make this specific team better, or whether the team needs a senior generalist more than a specialist this quarter. The cost of a wrong hire lands on the manager, which is why the decision stays with them.
Performance, feedback and conflict
Telling a strong engineer their behavior is hurting the team, managing someone out fairly, or mediating between two senior people who both have a point: none of this works delivered by a tool. Gallup's estimate that managers account for at least 70% of the variance in engagement across business units is the best-known number on why this half of the job matters.
Owning trade-offs and accountability
When an agent writes the code and another agent reviews it, someone still has to decide what "good enough" means and answer for it when it isn't. That's management work. On the teams I see doing this well, the engineering manager sets the AI code review bar and makes sure every agent-written change has a directly responsible individual. As AI output goes up, the demand for clear human ownership goes up with it.
Developing people
Career growth was already the part of management most often neglected, and AI makes it harder. If agents now do the well-scoped work that used to train juniors, someone has to design a deliberate path for people to build judgment. The same pipeline problem shows up in the junior developer data, and fixing it is a manager's job.
How does the engineering manager role change?
Wider, more technical and more explicitly accountable. Here's what I expect the job to look like on teams that adopt AI seriously:
- Larger spans. The small-team manager, with two or three reports, is the role the Google data targets. Expect fewer managers, each with more people.
- Humans plus agents. Microsoft's 2025 Work Trend Index calls this the "agent boss" and proposes a human-agent ratio as an org-design metric. For an engineering manager it means owning what the team's agents are allowed to do and how their output gets checked.
- More technical. You can't set a quality bar for agent-written code you can't read. Managers who drifted fully away from the codebase will feel this first.
- Less routing, more deciding. When the dashboard answers "what's the status?", the manager's value is in "so what do we do?"
- Ownership design as a core skill. Deciding who owns what, including which human owns each agent's output, becomes a daily part of the job, the same way delegation is for a CTO.
What should an engineering manager do about it this week?
Audit your own calendar. Take last week and mark every hour as either "moving information" or "making a call about people or trade-offs." If the first category is most of it, that's the part of your job a tool and a flatter org chart are coming for. The fix is to automate it yourself before someone else does, and spend the reclaimed time on the second category, where nobody has a replacement.
Then get hands-on with the agents your team uses until you can judge their output, and write down who owns what on your team, including the AI-written parts. If you're weighing whether the role is still worth it, the engineering manager salary data shows how the market currently prices it.
The verdict
AI will not replace engineering managers. It will replace the status-router version of the job, and organizations are already using that to run with fewer managers and wider spans. The manager who survives is the one whose week is mostly hiring, coaching, deciding and owning outcomes, including outcomes produced by agents. If that describes your job, the AI era gives you more leverage. If it doesn't, that's what to change first.
Sources
- NBC News: Google has eliminated 35% of managers overseeing small teams in past year, exec says (August 2025)
- Fortune: Amazon CEO Andy Jassy on the employee-to-manager ratio (November 2024)
- Gartner: Top Predictions for IT Organizations and Users in 2025 and Beyond (October 2024)
- Microsoft: 2025 Work Trend Index, The Frontier Firm is born (April 2025)
- Gallup: Managers Account for 70% of Variance in Employee Engagement
- Tech Manager Weekly #504 (19 October 2026), discussion of management in the age of AI.
Frequently asked questions
Will AI replace engineering managers?
Not as a role, but it is already removing a large part of the work and some of the positions. AI handles much of the status reporting, coordination and documentation that filled an engineering manager's week, and companies like Google and Amazon have used the resulting slack to flatten layers. The parts that remain, such as hiring, coaching, performance and owning trade-offs, are harder to automate and become a larger share of the job.
Will AI replace managers in general?
Gartner predicted in October 2024 that through 2026, one in five organizations would use AI to flatten their structure and eliminate more than half of their current middle management positions. That is a forecast about some organizations, not a measured outcome across the economy. The likelier pattern is fewer managers with wider spans, rather than teams with no manager at all.
Should engineering managers move back to an individual contributor role?
Only if you would want to anyway. Some of Google's removed manager roles reportedly became IC roles, so the path is real, but the IC job is also changing as agents write more code. A better test: if most of your week is status routing and meetings you could replace with a dashboard, either rebuild the job around people and technical judgment or move to IC deliberately before the decision is made for you.
Will engineering managers have bigger teams because of AI?
That is the direction the large-company data points to. When AI takes over status reporting and coordination, the overhead per report drops, and organizations raise span of control rather than keep the same number of managers. Wider spans only work if the manager also delegates ownership clearly, otherwise the manager becomes the bottleneck.
What skills should engineering managers build for the AI era?
Build technical fluency with the AI tools your team uses, so you can judge agent output and set the quality bar. Get sharper at the human work that does not automate: hiring, hard conversations, career development and conflict. And learn to design ownership, naming a directly responsible individual for agent-written code and AI-driven decisions so accountability does not dissolve as output volume rises.