You've been through this. Your team hasn't.
I remember the first time I went a whole day in management without shipping anything.
I’d been in and out of meetings, given some feedback in a 1:1 and dropped some pointers in a Slack thread. I closed the laptop at 6pm and couldn’t tell you what I’d actually done. Two months earlier I’d have spent the same day deep in a feature, watching tests go green, deploying to staging, finishing on a high. Now, I’d written a lot of words and done a lot of listening.
It felt unnerving. I kept reaching for the old proof—the PR I could point at, the dashboard I could deploy—and it wasn’t there. I assumed I was bad at the job. Eventually I worked out I was being introduced to the central feature of the job, which is that most of the meaningful work doesn’t leave a trace.
It took me about six months to make peace with that, and another year to stop treating it as a tax on real productivity and start treating it as the work.
I’d half-forgotten that wobble until last month, when I read the LeadDev State of AI-Driven Software Releases report and saw critical thinking ranked as the most important engineering skill for the next three years. Architecture, systems thinking, communication followed close behind. The bottleneck, the report concluded, has moved upstream. Coding is fast now. Framing the problem, decomposing it, deciding what done actually means—that’s where the real work has gone.
The report frames this as a soft-skills renaissance. Phew, the fundamentals still matter. Reassuring, if you’re a leader.
What I keep seeing in the coaching room is something subtler than reassurance, though: the wobble I went through fifteen years ago—sitting at my desk wondering whether words and judgments count as work—is now leaking out of the management track and into the engineers themselves.
Not identically, to be fair. Your engineers are still shipping. The PRs merge, the features deploy, the standup still has a what did you do column. They have tangible output, and tangible output is a perfectly reasonable guiding force for an engineering team—they’re not in management.
But more and more of their actual day looks like: reviewing code they didn’t write, debugging an assumption Claude made three steps back, choosing between four AI-generated approaches, framing a problem clearly enough that the agent doesn’t go wandering off. None of that lands neatly in a commit message.
The bit the LeadDev report doesn’t quite say out loud is that the ground is still crumbling under our feet. The rubrics most of us grew up on—the staff/senior split, the levelling docs, the company “what does great look like”—quietly assume a world where the work you most want from engineers is visible. Shipped features. Solved problems. Architectures drawn. AI doesn’t break those rubrics so much as make them measure a shrinking share of what actually matters. You can still reward visible output. You’re just rewarding a smaller slice of the iceberg every quarter.
Which is where engineering leaders have, surprisingly, an advantage I don’t see being named.
You’ve already done a version of this transition. Recently, in some cases. You know what it feels like to spend a Tuesday in meetings and end the day wondering if any of it counted. You remember the temptation to over-claim—to dress up “had a useful 1:1” as something more concrete just so you’d have something to point at. You remember the quiet relief when someone—a manager, a coach, a peer—finally named the invisible work as load-bearing rather than overhead.
Your team is now living inside a milder version of that experience and most of them don’t have a frame for it. They’re still trying to be practitioners while the practice is being quietly outsourced. The senior who’s gone a bit flat, the one who’s fine but oddly disengaged—that’s often where the wobble lands first.
The move isn’t to swap output for invisible work. The move is additive. Keep rewarding the visible. And start explicitly naming, valuing, coaching the upstream work that doesn’t ship—the framing conversation that saved two weeks, the let’s not build this call, the review where someone caught the silently-confident bug in Claude’s output. None of these appear in a PR, but all of them are increasingly where the leverage lives.
If you don’t name it, your strongest engineers will quietly reorient toward the visible—because that’s what gets rewarded—and lose the very judgement the moment is asking them to grow.
You spent some chunk of the last few years learning to stop being the best engineer in the room. Your team is now being asked, in their own way, to stop being practitioners.
You’re uniquely equipped to help them.