Nobody Will Count You If You Leave Early

On Sunday a software engineer, two and a half years into the job, posted a question in a forum where people ask about their careers.

Start with how they described their life. "At this moment in time, I'd say I'm very happy with my current job. Pay is good and so is the WLB." WLB is work-life balance. Nobody at their company has been laid off. They're 23.

Then the question. Should they leave the profession?

"If AI is inevitably going to kill this field, then I feel like I need to jump ship before it all sinks."

I read that post and thought about what has actually happened to them. Nothing. No cut, no reorganisation, no manager taking them aside, no software turning up that does their work. They're good at the job, they like it, they're paid well, and they're weighing up years and a pile of money to get out of it.

They aren't grieving a job. They still have the job. They're grieving a career, in advance, in a forum full of strangers.

And they aren't unusual.

In August, McKinsey published its yearly survey on AI: 1,719 people across 97 nations, from the full range of regions, industries and company sizes. Two of its findings sit near each other and point opposite ways.

Fourteen percent of respondents at organisations using AI say AI contributed to their workforce shrinking over the past year.

Thirty-nine percent expect AI to cut their headcount over the coming year. Last year that figure was 32%.

Those are not the same kind of fact. The 14% is a report: something happened, and they're telling you it happened. The 39% is a forecast: nothing has happened, and they're telling you what they think will.

Stay on the 14% for a second. Turn it around and roughly six in seven of those organisations reported no AI-driven fall in headcount at all. It simply didn't happen. If you walk into an average office tomorrow, the thing everybody is braced for has, so far, not arrived.

The gap between those two numbers is the space our 23-year-old is living in. It's also where almost everybody I talk to about this lives.

Picture a woman who runs a small shop. She hears a retail park is going up on the ring road. Nothing has been built. There's a planning notice and some talk in the pub. She's sensible, so she gets ready. She stops ordering the good stock, because why tie up the money. She lets the Saturday girl go. She doesn't bother repainting the front.

Eighteen months later the retail park is still a field with a fence around it. Her shop is finished anyway.

Not by competition. By eighteen months of behaving as though the competition had already arrived.

The forecast did the damage, even though the thing forecast never turned up.

For clarity, I should tell you where I'm standing. I'm one of the people a survey like that asks. I've sat in the rooms where a company decides what it thinks next year looks like, and I can tell you what it costs to tick "yes, I expect AI-related headcount decline" on a questionnaire in May.

Nothing. It costs nothing.

No budget is attached to it. Nobody has to be told. No severance gets modelled. An expectation is free. A decision is expensive, and it gets argued over line by line, and that is a large part of why the 14% is so much smaller than the 39%.

So the forecast runs miles ahead of the event, and it's the forecast that reaches the 23-year-old. They aren't reading the 14%. They're reading, in their own words, everyone recommending people "stay far away from the software engineering field", and they're making the only move they can see.

Which leads me to my next point.

If the employers in that survey are wrong, they lose nothing. They ticked a box in May and life went on.

If the 23-year-old is wrong, they've spent five years and a pile of money retraining out of something they were good at and happy in. They don't get the five years back.

We're asking the person with the least information and the most to lose to price a twenty-year forecast.

Plus, there's a counting problem underneath all of this, and I think it will look very obvious in a few years. Everything we measure is a separation. Companies report what they cut. Governments count who filed. But nobody counts the person who decided not to come in, or to leave early, or not to specialise, because of something they expect. That number doesn't exist. If it ever does jump, they won't show up in a single statistic about AI and jobs. They'll be a career change. They'll look like somebody who fancied a change of scene.

I'm not going to tell you they're wrong to be scared. The honest position is an uncomfortable one: the people best placed to know are telling surveys they expect cuts, and the same people are reporting that the cuts have mostly not happened, and both of those are true at the same time.

What I do know is that we've built a conversation about AI and work that talks almost entirely about the future and almost never about the present. The people absorbing it hardest are the ones with nothing to compare it against. They're in their 20's. They have never watched a labour market terrify everybody and then not do the thing.

The question is, what should we tell them?