You Cut Management and Compliance in the Same Sentence. Only One of Those Clocks Has Rung.

In 2010 I was working inside a subsidiary of a media holding company called Wahlstrom Group, on a project built jointly with a third-party vendor: technology that let major brands assemble their own advertising campaigns out of modular pieces, across whatever channels they wanted to run on. I was splitting my time that year between two managers. One of them, Joe, would turn out to be the best manager I had in thirty years of doing this work. The other reported through a CTO who was about to make a mistake that cost the entire subsidiary its existence inside of a week.

The CFO stopped me in the hallway one afternoon, excited. There was a demo coming, and he wanted to see the thing the whole project existed to produce: the finished ad, assembled, running. I mentioned the conversation to the CTO a few minutes later, expecting it to be useful information. He thanked me, then told me they were going to demo the backend instead: the system that builds the pieces, not the system where the pieces become something a brand could use. I told him the CFO wanted the finished product, not the infrastructure. He told me, in the way arrogant people do, that he knew what he was doing and that I was out of line.

The demo happened exactly as he'd planned it: genuinely impressive backend engineering, competently built and competently shown. Afterward the CFO pulled me aside. "What happened? That demo was a disaster." I told him the message had been delivered and dismissed. Within a week the project folded and the subsidiary closed.

I hate shelfware more than almost anything in this work: wasted effort that never got to matter, and I'd done everything inside my own ethics to make that project succeed despite what I could see coming. There was real relief when it ended, but it wasn't relief that the work was gone. It was relief that I no longer had to be in a room with that particular arrogance.

That CTO made a specific kind of mistake, and it's the same mistake a lot of companies are making right now, at much larger scale. Cloudflare's CEO wrote publicly this month that AI let the company cut a fifth of its workforce, and he named middle managers in the same breath as auditing, finance, legal, and compliance: functions the company no longer needed once AI absorbed the coordination work. Google cut over a third of the managers overseeing small teams. Gartner predicted, back in 2024, that a fifth of organizations would use AI to eliminate more than half their middle management by this year, and that prediction is arriving on schedule.

Management and compliance are the same category of function, whether or not the org chart says so: both exist to catch a specific class of failure before it compounds into something expensive. But they catch on different clocks, and the distance between those two clocks is what I call Clock Drift. Cut the management layer and the gap shows up fast: a sprint slips, a decision stalls, someone starts quietly absorbing coordination work nobody assigned them. It's visible within weeks, and because it's visible that quickly, it also gets corrected quickly, or waved off as a one-time hiccup. Cut compliance, audit, or legal oversight at the same time, and that gap runs on a slower clock: an annual audit, a regulatory review, or the first AI-agent decision nobody was positioned to own.

Here is what that produces. The fast clock rings first, gets resolved cheaply, and the organization reads its own quick recovery as proof the whole cut was safe: we removed the managers and nothing broke. That reading gets applied retroactively to the slow clock too, even though the slow clock hasn't come due yet, and a year of institutional confidence gets built entirely from the fast clock's evidence. Then the slow clock rings, usually at the worst possible moment, and it doesn't land as a second data point in the same pattern; it lands as a surprise, because everyone already decided the pattern was closed. The leaders who avoid this outcome aren't the ones who refuse to cut anything. They're the ones who already know which clock they're betting on.

AI is a genuinely useful tool inside a team that already has an architecture: a designed operating system with clear decision rights, cadence, and ownership. It is not a substitute for the catching function itself, no matter how fast it makes the execution underneath that function look. The question was never whether AI could do the coordination work faster. It was always whether removing the human who owned the judgment about when to catch something left anyone still positioned to catch it.

If you've cut, or you're considering cutting, a function you're calling overhead, ask one question before you decide the cut was safe: what has this function caught in the last year that never became a line item, and on what clock would you actually find out if removing it was a mistake? The CTO in that hallway was certain the technical achievement was the whole answer. He found out otherwise in a week. You might not get a week's warning.

I write about structural leadership for technical leaders in high-stakes operating environments. If you want to see where your system is load-bearing on you personally, the LeadershipOS™Scorecard maps it:https://theleadershiposbook.com/scorecard