A rollout can look finished on paper and be dead in the water. Earlier this summer I sat in a working session where someone described exactly that. The licenses were live, the dashboards were built, the training had been logged, and the team still would not touch the tool. Not because it confused them. When you got them talking, the reason underneath was plainer and more human than confusion. They had already done the math on what a fully working version of this thing meant for the number of chairs in the room. As one person put it, you can scrub the data as clean as you like, but if nobody will use the thing, you are spitting in the wind.

That instinct is turning up in the national numbers now, and it is not the story most executives are braced for.

The fear is that it works

A survey this spring of 2,400 knowledge workers across the United States, the United Kingdom, and Europe found that 29 percent of employees admitted to actively working against their own company’s AI strategy, a figure that climbed to 44 percent among Gen Z.1 The survey came from an enterprise AI firm with a direct stake in adoption, so take the framing with that in mind, though the behavior matches what managers have described to me for the past year. What matters is the reason. The most common one, named by 30 percent of the people doing it, was fear that the technology would take their job.1 People sabotage the tools they are most afraid will work. The refusal says nothing about whether the AI is any good. It is a bet that it is good enough to come for the person being asked to feed it.

When capable people work against a tool they believe works, they are protecting themselves, not doubting you. The meeting where the tool was announced was never going to hear that said out loud.

The ask nobody says out loud

Set that beside what the same survey found leaders planning. Sixty-nine percent of executives said they were already preparing AI-related layoffs.1 So the request on the table, stripped of its change-management language, is this. Help us stand up the system that is one reason your role may not exist in a year, and do it with enthusiasm, or we will note your reluctance. Framed that way, the sabotage stops looking like a discipline problem and starts looking like the most predictable thing in the building. You cannot order someone to be eager about building their own replacement. Pressure will buy you compliance in the room and resistance everywhere you are not looking.

The tools are a shakier bet than the mandate admits

I want to keep two things from collapsing into one, because they are not the same. The fear driving the sabotage is a separate matter from whether these tools actually deliver, and on that second question the record is not flattering. S&P Global Market Intelligence found that the share of companies abandoning most of their AI initiatives rose from 17 percent to 42 percent in a single year, with the average organization scrapping close to half of its pilots before they ever reached production.2 Most of those projects did not die because someone on the floor undermined them. They died of ordinary causes, a use case nobody had pinned down, a return that never arrived, a tool bolted onto work no one reorganized to receive it. Keep this separate from the workers’ fear. The company placing the bet is the one abandoning it, at scale, on its own.

The way you can lose both

Now run the two decisions together, because a great many companies are making both. You thin out the experienced people to help pay for the AI and to prove you are serious about it. Then, a year on, you land among the 42 percent who abandon most of what they built. You are left holding neither. The knowledge walked out with the people you let go, and the system that was supposed to cover for them is a decommissioned pilot. What remains is a short menu of unpleasant options. You can shrink what the company does until it fits the smaller crew. You can pile the missing work onto whoever survived and call it doing more with less. Or you can go back into the market to rehire the experience you just paid to remove, at a premium, having shown everyone still on the payroll exactly how little their tenure bought them. I said in that working session that a fair number of firms would spend the following year rehiring, once they found the tool could lift the routine load but could not own a decision or carry the accountability that rides with one. I have seen nothing since to change my mind.

Underneath all of it sits the same mistake I watched play out with Agile for the better part of twenty years. An organization reaches for a practice before it has named the problem the practice is meant to solve, then leans on pressure to close the gap the pressure itself opened. Ordered adoption is the weakest kind there is. A team that picks up a tool because it was told to will set it down the moment the order lifts, because the habit was never theirs. The version that lasts is the one where people take the thing up because it makes their work better and they can see that for themselves, and you do not get there by frightening the people you are counting on to carry it.

So here is the question I would put to any executive staring at a stalled rollout and reaching for a way to force the last holdouts into line. Before you do, ask what the people dragging their feet understand about this technology, and about how you mean to use it, that you have not been willing to say to them plainly. Then ask whether you are prepared to lose both the tool and the people if the honest answer is the one they already suspect.