8.4% of Us Say It Worked

I am the population of this survey.

Careerminds asked 600 HR professionals who had run layoffs in the previous year how it went. Not consultants, not journalists, not the people who were cut. The people who built the list, ran the meetings and signed the forms. My people, if you like, though I am not sure any of us would say it out loud.

Only 8.4% said the restructuring delivered exactly what was promised.

I have read that number several times now and it does not get smaller. More than nine out of ten HR leaders, asked in private about the thing they were paid to execute, could not say it delivered what they were told it would. In any other function that would be a scandal. In ours it is a survey result, published on a Tuesday, and the next wave is already being planned.

Let me show you the rest of the room, because the 8.4% is the headline but the other numbers are the story. Nearly a third of these leaders said their organisations lost critical skills and expertise in the cuts. A third had already rehired for between a quarter and half of the roles they eliminated, and another third had brought back more than half. Over half rehired within six months. Only about one in five said automation replaced the roles without operational problems, and four in ten said they would take a completely different approach next time.

Then the outside view, which lands in the same place. Forrester puts the share of companies that cut for AI and now regret it at 55%. Orgvue got to the same figure separately and did the sum: roughly $1.27 spent for every $1 saved, once severance, lost productivity and the cost of hiring the replacements are added up. Commonwealth Bank of Australia cut 45 customer service roles on the strength of a voice bot and reversed within weeks. Klarna, which went further than almost anyone on AI service, now has its chief executive saying there will always be a human if you want one.

I want to be careful here, because I am not against any of this. I use AI every day and I expect it to change how work is done more than anything I have seen in my career. The question I am asking is narrower and harder. If the people who executed the strategy say it did not work, what exactly was the knowledge industry taken apart for?

Here is how it looks from the seat. A restructuring on AI conviction does not begin with evidence. It begins with belief. Someone senior has seen a demo, someone else has a board deck with a productivity number in it, and the number arrives at HR already decided. Our job at that point is execution, and we are good at execution. We can take a business case that is really a hope and turn it into a list of names by Thursday. I have done it. Every experienced HR leader reading this has done it.

What we cannot do, from that seat, is make the hope true. AI either does the work or it does not, and no amount of execution quality changes which. This survey is 600 of us admitting that, in private, to a pollster, a year after the fact, when the people on the list are long gone and the requisitions to replace some of them are already open.

Now think about what the other side of that transaction looks like. Somebody spent twenty years becoming an expert at something. Whatever the reskilling industry says, a career is more than a skill set. It is the slow accumulation of judgement, the thousand things you know that never made it into a process document, the reason people came to you. That was taken apart on the strength of a belief. And the people who took it apart now say, at a rate of more than nine in ten, that the promise was not delivered as made.

They took apart the thing you spent twenty years building. 8.4% of them say it worked.

The reversals do not undo it, either. A third of these companies are rehiring, but a rehired role is a long way from a restored person. Whoever comes back, or whoever is hired in their place, is walking into a building that has proven, publicly, that it will price expertise at zero on a hunch. You cannot un-know that about the place you work. I suspect that is a bigger cost than the $1.27, and it is one nobody is counting.

That's the thing about this survey. Displaced workers complaining are easy to discount, and commentators have an angle. This is the executioners, reporting on the execution, and finding it wanting.

So what do we do with that? I honestly do not think the answer is "stop." AI is real and the pressure to use it is not going away. But a function that has just told a pollster, nine times out of ten, that the last round did not deliver what was promised has, at the very least, earned the right to slow the next round down and ask for the evidence before the list gets built. Whether we will use that right is a different question, and I am not optimistic, because the belief always arrives before the data and the list is always due on Thursday.

More than nine in ten of us could not say it delivered. I keep wondering how many of us said so in the room.