AI's Quiet Advance on White-Collar Work

Hey there! Welcome to the 5th issue of this newsletter, coming to you this week from France.

There's a new map of the knowledge economy. And most people haven't looked at it yet.

In early March, Anthropic published what might be the most detailed picture we have of how AI is actually reshaping work. Not theoretically. Not in five years. But right now.

Here's what you need to know and here is the link to the full report if you fancy a read -

https://www.anthropic.com/research/labor-market-impacts

What they measured and why it matters

Most AI-and-jobs research asks: what could AI theoretically do? Anthropic asked a different question: what is it actually doing, right now, in real workplaces based on real usage of their model, Claude?

The distinction matters more than it sounds.

They introduced a new measure called "observed exposure" or AI Displacement Risk — not what AI could automate, but what it is actually automating, weighted toward fully automated uses rather than human-assisted ones, tracked across 800 occupations.

What they found gives a sobering preview of what's ahead for white-collar workers. Theoretical AI coverage exceeds 80% in multiple occupation groups. Specifically, computer programming and math, engineering, legal, business, finance, and management.

But here's where it gets interesting. As we speak, actual coverage remains a fraction of what's feasible, due to legal constraints, technical hurdles, and the simple reality that humans still need to review AI's work. In line with what I covered in my last issue, Enterprise AI: Real Talk, the gap between reality and hype.

That gap is your window. But it is closing.

Who is most exposed — and it's not who you think

Here's the finding that reframes everything.

Workers in the highest-risk professions tend to be older, more educated, better-paid, and more likely to be women.

Let that sit for a moment.

This isn't a story about low-wage, low-skill workers being automated out of existence. The most exposed group is 16 percentage points more likely to be female, and almost twice as likely to hold a graduate degree.

The people who did everything right — got the degrees, built the careers, climbed into well-paid knowledge work — are disproportionately in the crosshairs.

Source: Anthropic Labour Market Impacts of AI Report, March 2026.

If you're reading this newsletter, there's a reasonable chance that includes you.

The entry-level alarm

There's a second finding worth flagging, especially if you manage people or are thinking about the pipeline below you.

The report argues that it found no systematic increase in joblessness among workers in heavily exposed occupations since late 2022 — but it did find suggestive evidence that hiring of younger workers has slowed in those same fields. This is already visible beyond the data.

The UK currently records some of the highest youth unemployment rates across EU economies — a trend we touched on two issues ago when we looked at how more jobs will change than disappear.

The deskilling question nobody wants to ask

If you delegate enough cognitive work to AI — the writing, the analysis, the synthesis, what happens to the sharpness that made you valuable in the first place?

It's a question Anthropic's research is beginning to surface. It doesn't answer it yet. But it's worth paying attention to.

For now, most people are using AI to think alongside them, not instead of them. But the direction of travel matters. The tasks being handed over are getting more complex — and the question of what happens to your own capability doesn't wait for full automation.

So what does this actually mean?

Anthropic built this index as an early warning system — designed to detect disruption before it becomes visible in unemployment data. The fact that they felt the need to build it tells you something.

But what the data means depends on how you read it.

Lens 1: The compression story

The signal isn't mass unemployment. Not yet. It's something quieter — a gradual narrowing of opportunity. Fewer entry-level roles. Slower growth in exposed professions. High exposure is concentrated precisely among the educated and well-paid.

In this view, AI doesn't replace entire jobs overnight. It compresses them. Fewer people are needed to produce the same output. And if you're a mid- to senior-level knowledge worker, you're not outside this shift. You're inside it.

Lens 2: The augmentation story

There's another way to read the same data.

AI is automating specific tasks — not whole roles. That opens the door to redesigned jobs, new workflows, and entirely new categories of work that don't yet have names. We've seen this before. The internet didn't eliminate knowledge work — it reshaped it entirely. In this view, AI becomes a layer that expands what individuals can do, rather than simply reducing demand for them.

Which lens are you looking through?

As always, your career isn't broken. The playbook is.

Gaziza

PS. One of the best ways to support this newsletter is to forward it to a colleague or friend navigating the same questions.

Was this forwarded to you? Subscribe here 👇.