AI is Already Changing the Labor Market. Anthropic Has the Data.

Anthropic published new research this week on AI's impact on the labor market. Not predictions. Not thought experiments. Actual data from how their models are being used in production, cross-referenced with labor statistics and task-level analysis of job functions.

The findings are more interesting than the headlines suggest. The usual narrative is binary: AI will either take all our jobs or it won't affect anything. Anthropic's data tells a different story, and it's one that founders and hiring managers need to understand.

What the Numbers Show

The core finding: AI is already handling a meaningful percentage of tasks in knowledge work, but it's not replacing roles. It's reshaping them.

Anthropic analyzed usage patterns across their enterprise customers and found that roughly 14% of all tasks performed by knowledge workers in their customer base have been partially or fully delegated to AI. That's not a projection. That's what's happening right now, today, in real companies.

But here's the nuance that matters. The 14% isn't evenly distributed. Some job functions have seen 40% or more of their tasks shifted to AI. Others have seen almost none. The distribution is lumpy and counterintuitive.

Writing and content creation: about 35% of tasks now involve AI assistance. No surprise there.

Data analysis and reporting: about 28%. Also expected.

Customer support: about 32%. This one's moving fast.

Software engineering: about 22%. Lower than the hype would suggest, but climbing.

Strategic planning and decision-making: about 4%. Almost untouched.

Sales and relationship management: about 8%. Humans still dominate.

The pattern is clear. AI is eating the production layer of knowledge work. The tasks that involve creating artifacts (writing documents, generating reports, writing code, drafting responses) are being augmented significantly. The tasks that involve judgment, relationships, and strategic thinking are barely affected.

What This Means for Team Structure

If you're a founder building a team right now, these numbers should change how you think about hiring.

The old model was to hire people primarily for their production capacity. You needed three copywriters because you needed to produce X pieces of content per month. You needed five analysts because you had Y reports to generate. You staffed based on output volume.

That model is breaking. AI can handle much of the raw production work. What it can't handle is the thinking that informs what to produce, the judgment about whether the output is good, and the relationships that determine who cares about the output.

I think the winning team structure for the next few years looks like this: fewer people, but each person is more senior. You want people who can direct AI, review AI output critically, and make judgment calls that AI can't. You don't want people whose primary value was producing volume.

This isn't a comfortable thing to say. It means fewer entry-level knowledge work jobs. It means the traditional pathway of "start as a junior analyst, learn the business, move up" gets disrupted. The junior analyst role was often mostly production work. That's exactly the work AI handles well.

The Skill Premium Is Shifting

Anthropic's data shows something else interesting. The economic value of different skills is shifting rapidly.

Skills that are becoming less valuable: the ability to write fast (AI writes faster), the ability to research efficiently (AI searches and summarizes faster), the ability to produce clean code for well-defined problems (AI handles this).

Skills that are becoming more valuable: the ability to evaluate AI output critically, the ability to frame problems correctly before handing them to AI, the ability to manage complex workflows where AI handles components but a human orchestrates the whole, and the ability to do things that require physical presence, relationship trust, or genuine creativity.

The last point deserves emphasis. "Creativity" as most people define it (combining known patterns in pleasing ways) is something AI does fine. Real creativity, the kind that produces genuinely new ideas and unexpected connections, is still distinctly human. But most jobs that claim to require creativity actually require pattern execution, and AI is eating that.

What Founders Should Plan For

Based on this data, here's what I'd recommend for founders thinking about team building and operational efficiency.

Audit your team's task distribution. What percentage of each role's time is spent on production tasks vs judgment tasks? Be honest about this. If someone spends 70% of their time on work that AI could handle, the role needs to be redesigned.

Invest in AI literacy across your team. The biggest productivity gaps aren't between companies that use AI and those that don't. They're between companies where every employee knows how to use AI effectively and companies where only a few people do. Make AI competence a baseline expectation, not a specialty.

Redesign roles around judgment, not output. Instead of hiring a content team that produces 50 articles a month, hire a smaller team that decides what 50 articles to produce, directs AI to draft them, reviews and improves the output, and measures what works. Fewer people, higher judgment load per person.

Plan for the transition honestly. Some roles on your team will change significantly. Some might not be needed in their current form. Having honest conversations about this now is better than layoffs later. Help people skill up into the judgment layer of their work.

Don't over-index on AI replacing everything. Remember: strategic thinking, sales relationships, and complex decision-making are barely affected. If your business depends on those skills, AI is a tool that makes your team more effective, not a replacement for it.

The Uncomfortable Truth

Anthropic's data confirms what many of us have felt anecdotally: AI is changing work faster than the public conversation acknowledges, but slower and differently than the hype suggests.

It's not coming for entire professions overnight. It's coming for specific tasks within professions. And the professions that adapt, that redesign themselves around what AI can and can't do, will be dramatically more productive than those that pretend nothing is changing.

The data is real. The transition is happening. The question for every founder is whether you're going to design your company around this reality or get surprised by it later.

I know which option I'd choose.