“Half the job got faster and half didn’t move at all.”

That’s how Tasha Ramesh, an Engineering Manager at Tinder, described managing engineers a year and a half into the AI shift. The operational side of the job, synthesizing information, surfacing blockers, tracking projects, has sped up. The human side, coaching engineers, giving hard feedback, resolving disagreements, hasn’t gotten any easier. If anything, it matters more.

Much of the industry has drawn a different conclusion. Companies are flattening organizations and asking managers to oversee larger teams. Meta’s Applied AI org runs roughly 50 engineers per manager . The implicit bet is that more capable engineers need less management.

That skips the more useful question: if engineers are becoming more capable, what kind of management do they need now?

We spoke with engineering leaders at Anthropic , Magnite , and Tinder , where I’m CTO, about what’s changing inside their teams. Four shifts stood out.

1. Managers need to become player-coaches

“Don’t track output. Track judgment.” — Aaron Wang, Engineering Manager at Anthropic

Now that engineers can generate code and explore solutions faster, sheer output tells managers less than it used to. Managers need to understand the thinking behind the work. What possibilities did you consider, and why did you choose this one? What trade-offs did you make? Why is this the right solution?

Managers can’t develop that judgment from the sidelines, which is why so many are taking on the role of player-coach. The best ones are building with AI themselves, testing the latest tools and forming their own view of what good AI-assisted engineering looks like.

2. Give engineers bigger problems and make them own the answers

“Managing senior engineers increasingly feels like managing engineering managers.” — Ryan Burns, Engineering Manager at Tinder

AI is giving individual engineers the capacity to do work once done by entire teams. As engineers direct AI themselves, managers have less reason to oversee every task. Their job shifts to setting the destination, providing context, and challenging decisions when needed.

There’s a paradox here. AI creates more autonomy, but autonomy makes good management more consequential. When an engineer can accomplish far more on their own, the stakes of their decisions rise with them.

That autonomy has to be real, though. If an engineer’s day becomes taking a ticket, handing it to an agent, and reviewing the code that comes back, the work has become a software factory.

“If you’re in a software factory where someone is just reviewing a robot’s code, I can see where burnout grows. My teams are working on novel things with agency, autonomy and architectural discretion.” — Bart Consedine, Head of AI Engineering at Magnite

The engineer is shipping more, but the job has lost what made it rewarding: solving hard problems, making decisions, building something themselves. That is its own form of burnout.

So the manager’s responsibility runs in both directions. Hand engineers the architectural decisions, and hold them accountable for the quality of what their agents produce. Ownership is what keeps the extra productivity from hollowing out the work.

3. AI can stretch a manager’s span, but only to a limit

“If a manager takes on too many direct reports, interactions risk becoming increasingly transactional.” — Aaron Wang

AI gives managers visibility. It can summarize projects, surface blockers, prepare for one-on-ones, and help a manager follow what’s happening across a much larger team. That’s why many companies are adding reports to each manager. Some are taking it too far.

AI might help a manager know what’s happening with 25 people. It doesn’t create the capacity for 25 coaching conversations about performance, ambition, conflict, or growth.

A better approach is to distribute some of that coaching. Senior engineers and tech leads can mentor closer to the work. Group coaching can spread lessons that would otherwise be repeated across one-on-ones. The structure has to keep managers responsible for making everyone on the team better and keep one-on-ones from turning transactional.

4. Faster teams need to talk more

AI makes engineers more independent, which raises the cost when they diverge. As execution accelerates, a week of work in the wrong direction costs more than it used to, and engineers spend more of their time on the strategic decisions where divergence starts.

That makes working sessions, real-time conversations, and in-person collaboration more important than before, not less.

Bart Consedine, Magnite’s head of AI engineering, spends roughly 30 hours a week in meetings and still protects time with his team: “I just have to have that dedicated time for [my reports]. That time is sacred.”

What managers are for now

AI is changing what engineers need from their managers: less coordination, more coaching, and a manager who has used the tools well enough to know good work when they see it.

As Tasha put it: “The more powerful these tools get, the more we need to connect with each other.”