What running performance for 16,000 managers taught me about AI

Across the five years I spent running performance for more than sixteen thousand managers, inside a decade at a Fortune Global 500, I heard a version of the same sentence at every level of the organization. From the executive floor down to a first-time team lead. The sentence was, roughly, the system is not working, and it is someone else's fault.

That decade taught me something that has turned out to be the most useful thing I know now that AI has arrived. Most of what looks like a people problem is actually a design problem wearing a people problem's clothes.

When a manager runs their team into the ground every quarter, the easy story is that the manager is bad. Sometimes that is true. Far more often, the manager is operating inside a process that was designed badly, with evaluation criteria that reward the wrong thing, decision rights that are unclear, and no capability ever built for the job they were actually promoted into. The manager is not the root cause. The manager is the place the root cause becomes visible.

I bring this up because AI has walked straight into the same trap, at scale, across the entire market, in about eighteen months.

Here is what AI does, reliably. It makes whatever is already there more visible and faster. If your managers were already strong, AI compounds that strength. If your managers had real capability gaps, and most do, because almost nobody was ever trained to lead inside the work they were handed, AI does not fix that. It makes the gaps louder. A manager who could not give clear feedback before now cannot give clear feedback faster, with an AI tool generating drafts of feedback they do not know how to stand behind.

This is the part that the technology-side conversation keeps missing. I sit in an unusual seat. I came up through learning and organizational effectiveness, the people side, but I have spent my career next to the technical implementations, and right now I am building custom GPTs to redesign real processes. I hold both rooms. And what I can see from that seat is that AI adoption keeps getting handed to whoever owns the platform, usually IT, and the people side gets called in last, after the rollout, to drive adoption. By then the design decisions are made and the managers are already improvising in fifty directions.

AI will not make a poor people manager a better one. I want to put that as plainly as I can. It will make the gaps more visible, and it will make the well-designed parts faster. Both of those are useful, but only one of them is what most companies think they are buying.

I did not learn any of this from a credential, though I hold the credentials. I learned it being accountable for it, at scale, on real Monday mornings, with real managers and a real budget and a board that wanted results. That is the lens I bring to a mid-market company now. Not a framework I read. A pattern I ran.

If your AI is live and your managers are stretched and the outcomes are flat, the problem is almost certainly not the tool. It is that the work was never redesigned and the capability was never built. That is fixable. It is the work I do.