Ask me what I'm proudest of from fifteen years at MED-EL and the expected answers are all on my CV: a global IT organization, the company's first AI strategy, ISO 27001. The answer I actually believe is less quotable. It's the stretch, some years in, when I could be away for weeks and nothing escalated. Not because nothing broke - things always break - but because the people who owned helpdesk, operations, security, cloud and infrastructure, governance and strategy handled it the way we had agreed things get handled, without checking whether I would have decided the same.Those managers were the actual infrastructure. Everything else ran on them.
The escalation point doesn't scale
I grew into that organization from a helpdesk seat, so for years my instinct was to be useful the way I had always been useful: know the systems, take the hard ticket, be the escalation point. That instinct caps an organization at the size of one person's attention. A hundred people across fourteen countries cannot run through one inbox.So the job changes. What each level owns has to be worked out, written down, and then defended - mostly against yourself. Every time I reversed a manager's call because I would have made it differently, I wasn't correcting a decision. I was announcing to everyone watching that the decision rights were fiction. A manager who has never made a call I disagreed with isn't leading. They're forwarding.None of that happens on goodwill. You hire for it, you make the standards explicit enough that someone who isn't you can hold them, and you hand over real authority in public, where taking it back would cost you something. Then you find out whether you meant it.
The hire I watched not work
This is also where I made my most expensive mistake, the one I still see engineering leaders make constantly: I gave a team to the strongest technical person in the room.On paper it was obvious. Nobody knew the domain better. Nobody had more credibility with the people doing the work. Within months it was just as obvious that expertise was the only thing being led with. The team slowed down. Decisions queued at one desk. The standard I thought I had delegated came back to me one escalation at a time.The mistake I own is not the hire. Selection is probabilistic; sometimes you're wrong. The mistake I own is the waiting. I saw it early and acted late, because the person was genuinely excellent and I kept believing the leadership side would catch up with a bit more coaching, a bit more time. It didn't, and in the end we parted ways - later than was good for the team, for the function, or for the person, who deserved clarity sooner than I provided it.Two things changed after that.I stopped hiring managers for technical excellence. It still matters, but it's the entry ticket, not the job. What I probe now is the management itself: how they built the people under them, what standard they held when it was inconvenient, what they did the last time delegation went wrong on their watch. If every story is about their own output, I have my answer.And I gave myself a rule for the waiting, because I now knew my failure mode was patience dressed up as fairness: the moment I catch myself quietly doing parts of a manager's job, the decision is already made. The only question left is how long I'm willing to pay interest on it.
What "running without me" looked like
The proof was never dramatic. It was routine.Certification audits ran with my managers carrying them - owning the evidence, facing the auditors - while I did the executive-sponsor part: mandate, investment, and the occasional cross-functional fight that needed my level. Budget decisions and hires happened inside frameworks we had agreed on, and I heard about a good share of them afterwards. That is how it should feel. I could be absent, properly absent, weeks at a time, without a single escalation that couldn't wait for a Tuesday.I want to be precise about what that evidence is and isn't. When I left in 2025, I didn't get to run the experiment people imagine; nobody gets a clean control group for their own absence. What I can verify is what I saw while I was still there: an organization that had stopped needing me day to day, run by managers who held their functions like their own. That is the only version of legacy a leader can actually check. Everything after your last day is somebody else's story, and claiming it would be exactly the kind of unfalsifiable talk I distrust in everyone else.
The part AI didn't change
I've spent the time since building a commercial unit at SQUER and watching enterprises try to adopt AI. Different work, same conclusion. Whether the initiative is an ISO program, a cloud migration, or an agentic-AI rollout, it stands or falls on whether there are managers who own their piece, hold a standard without escalating it, and make good calls when nobody senior is in the room.Strategy decks don't create those people. Tooling doesn't either. They get built slowly, through who you hire and what you genuinely let them own. That is why so few organizations have them.The model was never the hard part. The org chart was.
