For most of my first decade as a manager, I measured myself by what I added. The thoughtful comment on the design doc. The careful rewrite of a brief that wasn’t quite right. The Slack message that landed at exactly the right moment in exactly the right thread.
Somewhere in the past year that scoreboard has stopped making sense to me. The work that actually moves the needle is increasingly the work I don’t do. Briefs I don’t over-specify. Reviews I don’t request. Conversations I don’t hold because the team is already holding them.
This isn’t a story about absence as a strategy. The lazy version of “step back” is a different post. The harder thing underneath is that subtraction only works when the scaffolding around it is solid: clear end states, written standards, accountability that doesn’t depend on me being in the room, protection from the parts of the organisation that would punish independent judgement if I weren’t standing in the way. Without that scaffolding, stepping back is just abandonment with a leadership vocabulary attached.
The end state is the only thing you get to write
A useful way to test whether you’ve actually delegated something is to wait for an external party to put unmovable pressure on the team and see what happens. The first time it landed on one of mine and I genuinely stayed out of the way, my entire contribution was a single sentence: the constraint is real, you decide how we hit it. By the following day the team had reshaped the timeline, drafted a plan I hadn’t seen, and committed to it publicly. None of it was how I would have done it. All of it was better than what I’d have produced.
The principle is easy to say and difficult to live: write the end state, name the constraints, then stop talking about the path. A well-formed end state is a contract, not a brief. It says what must be true at the finish line, in terms specific enough that a stranger could check it. It says nothing about migration order, branching strategy, or feature flag mechanics. The moment I start writing those, I’ve stopped describing the destination and started writing the team’s design doc for them.
The hardest part isn’t the writing. It’s the silence between the brief landing and the plan coming back. Every instinct wants to fill that space: drop a “sense-check” comment, suggest a direction, appear in the conversation in the way that’s most visible. Most of those interventions don’t help. Most of them quietly tell the team that their thinking is on probation and my approval is the real contract, regardless of what the written brief says.
If it isn’t written, it isn’t a standard
The thing that sits next to a clean end state, and makes one possible, is a written baseline of what good looks like. Not aspirations. Not a manifesto. The non-negotiables.
We learned this the embarrassing way in a meeting about training budgets. Five engineering leaders and one person from L&D, trying to work out why a junior had been given contradictory answers about conference eligibility. Halfway through, the L&D lead pulled up the actual policy. None of the rules we’d been confidently citing were in it. They were preferences someone had stated in passing years earlier, relayed through Slack and 1:1s until they felt like institutional truth. The original context was long forgotten. The “rule” was being enforced by people who had inherited it as a wall.
Once I saw the pattern, I started seeing it everywhere. Engineering “standards” that were really one architect’s preferred way of working presented with the weight of policy. Review processes whose enforceability depended on which senior happened to be in the room. Documents long enough that a sensible engineer would read the first page and start guessing about the rest.
When the written baseline goes quiet, authority fills the gap. The same exploratory work gets celebrated from one direction and triggers a governance review from another, not because anyone is being deliberately unfair, but because without a documented standard every decision about what needs approval is made on vibes. If a stranger can’t extract the rule from the artifact, the rule isn’t really there.
The fix is unglamorous. A small number of things, written down, short enough that everyone reads them and strict enough that nobody escapes them. Nothing about culture. Nothing about values. Things you can check against an artifact, yes or no. That kind of document is what makes “define the end state” portable across teams and quarters. Without it, the end state gets re-litigated in every room by whoever happens to be senior that day.
Coaching without shielding is exploitation
The next lesson was harder, because it looks like the opposite of subtraction.
Twice in the past year I’ve watched capable team leads come close to resigning. Both had been on the receiving end of months of deliberate coaching from me: how to delegate, how to push back, how to set constraints instead of methods. Both were objectively better leaders than when we’d started. Both were absorbing organisational dysfunction faster than they could replenish.
The mistake was mine. I had been doing the developmental half of the job thoroughly and the protective half almost not at all. Teaching them to take ownership without changing the conditions under which ownership was being punished. Teaching them to delegate without addressing the culture that made delegation feel like a career risk. Building muscle in a gym that was structurally on fire.
Protective work is hard to see and impossible to celebrate. It looks like a difficult conversation upward about a stakeholder’s behaviour. It looks like declining to forward an unreasonable request that would have landed on the team in another half hour. It looks like absorbing political damage so the team gets to keep operating in the space the brief promised them. The outcome is the absence of a problem, not the presence of a win.
But this is the half that makes the rest real. If you coach a lead to step back into a system that will punish them for stepping back, you’ve taught them a skill the organisation will use against them. If you tell a team to “decide how we hit it” while the surrounding culture treats every imperfect choice as evidence the team shouldn’t have been trusted, you’ve handed them autonomy as a trap. Development and shielding are one job. Pick one and it stops working.
Watch the silence, not the output
Subtraction has a failure mode that doesn’t show up in the obvious metrics, and it took me longer than I’d like to admit to notice it.
There was a week earlier this spring when, by every visible measure, my teams were operating beautifully without me. A leadership transition handled cold. Cross-cutting quality work picked up from the bottom up. A senior stakeholder’s heavy scrutiny on a complex design absorbed and answered without anyone escalating in my direction. I went looking for somewhere I’d been asked for input that week. I couldn’t find one.
I almost wrote the satisfied version of that story before I noticed the other half. The same teams generating the visible wins were quietly slipping on the work nobody was watching. A commitment had moved two sprints without a clear reason. A cross-team integration had gone yellow without anyone naming it. An expensive piece of contract debt was being deferred for the third quarter running, because nobody celebrates the team that cleans up someone else’s mess.
Capability had shown up where we were watching. Accountability had gone quiet where we weren’t. Because nobody was watching, the pattern had nowhere to land.
When you build teams that don’t need you, management stops being about the bright edges of what’s getting done and starts being about the dark edges of what isn’t. The signals that matter aren’t in the celebrated PRs or the heroic deadline saves. They’re in the line items quietly slipping at the bottom of a roadmap nobody opened this week. Subtraction works only once “be in the loop” is replaced with “watch the loops nobody else is closing.”
Presence offers safety. Absence offers permission.
The last piece, and the hardest to internalise, is what your absence means to the people you manage.
When I’m in the room, or the channel, or the calendar, my leads know there’s a backstop. If the conversation goes sideways, I’ll help. If the decision goes wrong, I’ll course-correct. That safety is genuine and I spent a long time building it. It is also, increasingly, the thing that keeps the most mature leadership work I’ve trained them to do from actually happening.
I noticed this the first time I took an unstructured day off and went genuinely dark. Three different leads, on three different teams, independently did the most senior thing I’d seen any of them do. One wrote a formal message to a peer naming unprofessional behaviour, citing the standard, and asking for a written follow-up. One opened a difficult message to her own team with a public acknowledgement that part of the problem was hers. One produced a piece of customer analysis any principal engineer would have been proud of, without a prompt or a template. None of it was for my benefit. The only thread connecting the three was that I wasn’t there.
Presence gives safety. Absence gives permission. Both are real. Both are necessary. The conversation about engineering leadership focuses almost entirely on the first and almost never on the second. In practice this has changed a few small things. I decline meetings I used to attend out of “just in case” anxiety, and most of them turn out to be fine. I’ve stretched the cadence of some 1:1s deliberately, because the most interesting growth shows up in the space between sessions. And I pay closer attention now to what happens during my absences than during my presence, because that’s where I get the truest signal about whether the scaffolding underneath is real.
The job, restated
I started writing this thinking it was five separate ideas. Halfway through I realised it was one idea wearing five outfits.
Define the end state and let go of the path. Document the standards that make letting go possible. Shield the people you’ve coached, so the autonomy you’ve handed them is real rather than performative. Watch the dark side of the road as carefully as the bright side. And when the infrastructure is there, actually leave the room.
The discipline is to leave the space empty on purpose, and to spend the time I get back on the things only I can do: the contract above the team, the documented baseline beneath it, and the political weather around it that nobody else is going to manage.
I’m not particularly good at this yet. I still rewrite briefs I should have left alone. I still drop into threads I should have stayed out of. I still feel the involuntary jolt when a team makes a decision I would have made differently. But the direction is settled. The subtraction job is the job now. The rest is learning to trust it.