Anthropic Just Published the Data on Which Jobs AI Is Actually Eating. Here's What It Means for You.

Everyone's asking the wrong question.

"Will AI take my job?" is the wrong frame.

The right question is: are you being hired into a role that's already closing?

On March 5, 2026, Anthropic published their labor market impact study based on real Claude usage data. Not predictions. Not models. Actual usage data from the people who use their AI most. The findings are more specific than the headlines suggest.

What the data actually shows

Anthropic's study identified the job categories with the highest AI exposure based on real usage patterns:

1. Computer programmers and software developers (75% exposed)
2. Customer service representatives (70%)
3. Data-entry keyers (67%)

The study also named financial analysts and market research analysts among the highly exposed roles.

These aren't abstract projections. They're roles where the actual work is already being done, at scale, by AI tools.

But here's the nuance most coverage missed.

The disruption isn't mass layoffs. It's a hiring freeze.

Anthropic's data showed that the job-finding rate for workers aged 22 to 25 in high-exposure roles dropped by roughly 14 percent relative to 2022. The people in those roles now aren't being fired. But the next generation trying to enter those roles is finding a closed door.

A parallel data point from Harvard Business School research (February 2026): routine job postings fell 13% after ChatGPT's release. At the same time, demand for analytical and creative roles grew by 20%.

The door is closing on one side of the labor market. The other side is expanding.

Why this pattern matters more than the layoff narrative

The mass-layoff story is loud and emotional. It's also the wrong thing to be afraid of.

If you're currently employed in a high-exposure role, you probably won't be fired next month. Companies are cautious about that. The PR risk alone makes it complicated.

What's actually happening is quieter. New graduates in software development, customer service, data analytics, and financial analysis are finding fewer entry points. The pipeline that used to convert degrees into jobs is narrowing.

Fortune's reporting on the study described it as a "Great Recession for white-collar workers" - not because of mass unemployment, but because the entry-level pipeline is drying up.

There's a structural problem building here.

Mid-career professionals in high-exposure roles are relatively protected today. But in five years, when organizations have adapted their workflows fully, the protection disappears. If you haven't built the skills and the reputation to move laterally or upward, the options narrow.

The other side of the data

According to PwC's 2025 Global AI Jobs Barometer, roles requiring AI skills command a wage premium of around 56 percent.

That gap is not a function of companies paying for novelty. It's a function of output. Someone who can use AI effectively produces substantially more than someone who can't. The premium tracks the productivity difference.

The Harvard Business School data adds another layer: demand for analytical and creative roles grew 20% post-ChatGPT. The jobs that require judgment, synthesis, relationship management, and contextual reasoning are expanding.

The combination looks like this: the routine is contracting. The contextual is growing. The people who bridge both earn the premium.

What to do with this

1. Check your role against the exposure categories.

Don't do this defensively. Do it strategically. If your core work involves tasks that an AI can replicate at 80% quality in a fraction of the time, you have a window. The window is not permanent.

The question isn't "is my job safe?" It's "what percentage of what I do each week could be done by an AI tool today, and what's left?"

The answer to what's left is your actual value in the next labor market.

2. Identify one AI skill that applies directly to your work in the next 30 days.

Not a course. Not a certification. A specific application.

If you're in marketing, it might be building an AI-assisted content pipeline. If you're in finance, it might be using Claude to automate your reporting workflow. If you're in customer service, it might be learning to manage an AI-first support system rather than being part of a human-only team.

The skill you need is the one that moves you from replacement risk to deployment architect.

3. Build your personal brand around domain expertise plus AI fluency.

This is the combination that commands the salary premium.

Domain expertise alone is table stakes. AI fluency alone is temporary - it will be commoditized within 18 months as tools get easier. But the person who is the recognized expert in their specific domain AND who can use AI to operate at scale... that person is not exposed.

That's the profile that organizations will pay a premium for in 2026 and beyond.

The data is in. The pattern is clear. The window to move is now.

Related: how Jackson runs AI agents as an executive team and work with Jackson on AI systems.