Two things are true in 2026, and they contradict each other. The first: the most powerful people in technology keep saying your job is about to vanish. Anthropic's chief executive warned that AI could wipe out half of all entry-level white-collar roles within five years. Microsoft's AI boss put a clock on it, predicting "human-level performance on most, if not all, professional tasks" inside 18 months. A whole genre of viral "white-collar bloodbath" essays has grown up around the fear.

The second thing that is true: by the account of the people actually running companies, AI has barely done anything yet. I lead AI inside a large enterprise, so let me say the quiet part plainly. The layoffs are real. The robot's role in most of them is mostly a press release. The gap between those two truths has a name — AI washing — and understanding it is the difference between managing your career on facts and managing it on someone else's marketing.

The predictions got very loud

It is worth being fair to the forecasters: the direction is right. Generative AI genuinely automates real work, and a serious labour adjustment is coming. The problem is the calibration. "Half of jobs in five years" and "all professional tasks in eighteen months" are not forecasts; they are vibes with a number attached, and they are repeated because fear travels further than nuance. Even Sam Altman, hardly an AI pessimist, has started pouring cold water on the jobs-apocalypse framing — which tells you the most aggressive predictions have outrun what their own makers will defend.

The data got very quiet

Now hold those predictions against the evidence. A 2026 National Bureau of Economic Research study of executives across four countries found that about 90% of managers reported no impact on employment at their organisation over the past three years, and roughly 89% saw no change in productivity. A quarter of the senior executives surveyed don't use the technology at all. This sits on top of the now-infamous MIT finding that 95% of corporate AI pilots delivered no measurable return.

Read those two sections back to back and the contradiction is glaring. The companies cutting jobs "because of AI" are, by their own managers' testimony, not yet getting much from AI. So what is actually doing the cutting?

Meet "AI washing"

Here is the uncomfortable mechanic. After years of pandemic-era over-hiring, higher capital costs, and a cautious macro climate, a lot of firms wanted to shrink anyway. AI arrived as the perfect cover story. "We are restructuring because we over-hired and rates are high" is a confession; "we are leaner because of our bold AI transformation" is a strategy the share price likes. Harvard Business Review documented the tell: companies are increasingly laying people off for what AI might do, not what it has actually done. The practice even has its own label now — "AI washing", dressing ordinary cost-cutting in transformational clothes.

That does not mean none of it is real. When Salesforce cut around 4,000 support roles after AI agents began handling about half of customer conversations, that is genuine task automation showing up in a headcount line. But notice how rare the specifics are. For every Salesforce that can point to a measured deflection rate, there are a dozen press releases that wave at "AI efficiencies" without a single number — because the number would not survive contact with that NBER table. AI isn't taking your job. It's giving your CFO a better word for a decision they had already made.

What AI is actually automating in 2026 (the honest version)

Strip away the theatre and there is a real, narrower story, and it is the one worth planning around. AI in 2026 is coming for tasks, not titles. The distinction is your whole career.

  • What it genuinely does now: high-volume, well-bounded, repeatable work — routine support tickets, boilerplate and glue code, first drafts, high-frequency HR and admin queries, summarising and reformatting. Where a job is mostly a stack of those tasks, it compresses fast.
  • What it still cannot do: judgment under ambiguity, accountability for a decision, navigating organisational politics, original problem framing, earning a client's trust, and anything where being confidently wrong is expensive. Where a job is mostly those, it is barely dented.

Most real jobs are a blend, which is why "your job" is the wrong unit of analysis. The right question is: what fraction of my week is bounded task-work a model can already do, and what fraction is judgment and ownership it cannot? That ratio, not a CEO's round number, tells you how exposed you are.

The real damage is at the bottom of the ladder

If there is a part of the panic that deserves more attention rather than less, it is this. The tasks AI does best — the bounded, repeatable, "learn the ropes" work — are exactly the tasks we have always handed to juniors. So the squeeze lands first on the entry level. Yale researchers find the real job destruction is hitting before careers can even start: graduate and entry roles are frozen or thinned, because they are the cheapest to defer and the easiest to rationalise as "automated."

This is the genuinely dangerous trend, and it is the opposite of the headline one. The risk is not that 50% of today's workers vanish overnight. It is quieter and more corrosive: you cannot grow the senior talent of 2032 if you stop hiring the juniors of 2026. A company that AI-washes away its entry rung is eating its own seed corn — and won't feel the hunger until the pipeline it dismantled is the one it needs.

My take

I sit on the side of the table that deploys this technology, so I will be blunt with both sides of the room.

To leaders: stop AI-washing. Your best people can read an NBER table too, and the fastest way to lose them is to insult their intelligence with a transformation story your own dashboards do not support. If you are cutting because the business needs to shrink, say so. If you are betting on AI's potential, say that — and then actually resource the redesign instead of just the redundancies. Credibility is the one asset you cannot re-hire next quarter. And protect the entry rung, or budget now for the leadership vacuum you are creating.

To everyone worried about their job: the panic is overblown, but the shift underneath it is absolutely real, and the response is the same either way. Move up the ladder from doing tasks to judging, supervising, and owning outcomes — the work that AI-washing cannot fake and agents cannot yet shoulder. Learn to direct AI rather than compete with it. The people who thrive in this decade will not be the ones who feared the robot; they will be the ones who saw clearly that, in 2026, the robot was still mostly an excuse — and used the time that bought them to become genuinely hard to automate.