Let’s talk about the layoff that had no author.
There’s a memo that goes out most weeks now. It doesn’t usually say the words. It says restructuring. It says the way we work has changed. It says we’re leaning into automation so we can serve you better, and it’s signed by a person nobody has met.
And four hundred people find out that the thing that replaced them is a system that can’t take a call from an angry client at four on a Friday, can’t tell when a deal is quietly dying, and can’t be fired.
Now, the example that keeps this honest, because I’m not here to argue a slogan. Klarna. The buy-now-pay-later company. In 2024 it announced its AI assistant was doing the work of 700 customer service agents, and it was loud about it. The market loved it. It became the proof case for the whole category. Then the reviews came in, customers wanted a human, and the company went back to hiring people, and its own leadership admitted the quality wasn’t there.
Nothing about that is a scandal. That’s a company finding out what automation actually does. What matters is who got told what, and when, and who took the credit.
So let’s get the mechanism right, because the pattern is not AI bad.
Companies are expensive, and salaries are the biggest line on most income statements. Any CFO under pressure has always wanted fewer of them. AI didn’t create that desire — it was there in every downturn going back to the beginning of the corporation. What AI changed is the justification.
For thirty years, cutting heads meant admitting something. Demand fell. We over-expanded. We bought the wrong thing. The plan was wrong. There was shame attached, and shame had a name on it. Somebody owned it.
The machine arrived and handed the board a story with nobody’s fingerprints on it.
We didn’t get it wrong. We got automated. And the market pays for that version. Announce a ten percent cut for efficiency and the share price wobbles, because efficiency sounds like demand is soft. Announce that you’re reshaping the business around AI and it goes up, because that sounds like margin. That’s the entire incentive. The machine is being used as an alibi, and the alibi is one of the few things left in this economy with a positive return.
What people are missing is what the alibi costs the person holding it.
The moment you announce that a machine replaced your staff, you’ve told every customer that your work is machine work. You have said, in a press release, that what you sell can be produced by software you don’t own, running on a model someone else trained, priced by someone else’s meter, and changed by someone else’s product roadmap. Every competitor now knows exactly what to undercut. Every buyer now knows your service is a template. And inside the company, the people who remain learn the actual lesson — that the fastest way to be safe is to look irreplaceable and say nothing — which is the precise opposite of the honesty you’ll need in eighteen months.
The second cost is worse and it lands on the people walking out the door.
Losing a job is brutal in the ordinary way. The money, the year, the school fees. Being cut because the company needed a smaller payroll is at least legible. You can carry a decision. Being told you were replaced by a system that cannot do your job — that your judgment was worth less than a subscription — is a different injury, and most of the time it isn’t even true. You can’t argue with it, because the thing that supposedly beat you isn’t a person. You’re not being outworked. You’re being edited out of the story, and you’re expected to be dignified about the edit.
Now flip it, because the same alibi is running in the other direction too.
Bad decisions are increasingly attributed to the machine instead of the person. A model recommends a denial and the bank hides behind the recommendation. An agent says something the company never meant and the support queue shrugs it off as the bot being confused. Except that trick doesn’t survive contact with a tribunal. Ask Air Canada — a tribunal there held the airline responsible for what its own chatbot told a passenger about a bereavement fare. The company tried the alibi. It lost. The machine doesn’t sign for anyone. It never has, and no amount of confident language will give it a bank account.
Both sides of the pattern are the same move: a decision that belongs to a person, issued in the passive voice.
Here’s the operator part, and this is the piece I’d want my own team to hear.
If you lead anything, keep the author on the decision. Two rules.
First: the person who made the call signs the message. Not the business. Not leadership. A name, a reason, and the part you got wrong. That’s not decency for its own sake — it’s the only version of the story your remaining people will believe, and belief is what a company actually runs on once the easy years are gone.
Second: never blame the machine for a call you’d be embarrassed to defend with your name attached. If you wouldn’t say I decided this, you haven’t made a decision. You’ve found a scapegoat that can’t resign, and everyone in the room can see it.
For everyone else, the read is colder. When you hear that AI made the cuts, ask the question that has no good answer. Which roles, specifically. Which process. What did the before look like. Who owns the after. Most of the time the honest answer is that the seat was cut for cost and the machine is set dressing. That answer tells you exactly how this company will treat you at the next turn. Use it.
And if you’re selling — if you’re a founder standing in a room where a client proudly tells you they replaced their team with a model — understand what you’re actually being offered. You’re not being hired as the tool. You’re being hired as the alibi. The day it goes wrong, the sentence in the meeting will be that the AI did it, and the AI has no lawyer, no balance sheet, and no name. Guess who does. Price the work for that, or don’t take it.
The principle worth defending here is small and old. A decision needs an author.
And that’s what accountability actually is — not a policy, not a review board, not a log file. It’s a person who can be asked why. Who was in the room when it mattered. Who has to live with the answer afterwards. Every time we let the machine stand in for that person, we make the decision cheaper than the human it lands on.
You can automate the work. You cannot automate the responsibility for it.
So when the memo comes, do what you’d do with any document. Find the subject of the sentence. If there isn’t one, you’ve just learned everything you need to know about the people who sent it.