Why this matters for you: This is not a think tank prediction. This is not a government forecast. This is the company that builds the AI telling you, with its own data, exactly which jobs its technology is replacing — and which ones it cannot touch. If you work in fitness, hospitality, or wellness, this report is the most important validation of your industry’s future ever published.
What Anthropic Found
On 5 March 2026, Anthropic published “Labor Market Impacts of AI: A New Measure and Early Evidence,” authored by economists Maxim Massenkoff and Peter McCrory. The paper introduces a new metric called “observed exposure” — measuring not what AI could theoretically do to a job, but what it is actually doing right now, based on real-world Claude usage data across hundreds of occupations.
The distinction matters enormously. Previous studies estimated which jobs AI might affect. This study measures which jobs AI is affecting.
The Numbers
The Untouchable 30 Percent
Here is the finding that should make every gym owner, fitness instructor, personal trainer, and leisure facility manager sit up straighter: 30 percent of the entire workforce has zero AI exposure.
Anthropic’s own data shows that the following occupations have no measurable AI task coverage whatsoever:
- Cooks
- Bartenders
- Lifeguards
- Fitness trainers and instructors
- Motorcycle mechanics
- Dishwashers
- Dressing room attendants
These are jobs that require physical presence, human interaction, real-time judgement in physical environments, and — critically — the kind of face-to-face connection that no large language model can replicate. They are inherently human roles.
This is not a prediction. This is measured, observed reality from the company that builds the AI.
The Gap Between What AI Could Do and What It Is Doing
Perhaps the most striking finding in the entire paper is the chasm between AI’s theoretical capability and its actual deployment:
Across all occupations studied, 97 percent of tasks observed in Claude’s usage data fall into categories rated as theoretically feasible for AI. The technology can already do almost everything. Adoption just hasn’t caught up yet.
As the researchers put it: “The red area of current adoption is dwarfed by the blue area of what’s possible.”
When that blue area turns red — when businesses adopt what the technology already offers — the displacement will be sudden and enormous.
Who Gets Hit Hardest
The demographic profile of the most AI-exposed workers challenges every assumption about who technology displaces. This is not blue-collar vs white-collar. The most vulnerable workers are:
The workers with the most to lose are the educated, well-paid, white-collar professionals who thought automation was something that happened to factory workers. Graduate degree holders are four times more represented in the most-exposed occupations (17.4 percent) compared to unexposed ones (4.5 percent).
The Young Are Already Feeling It
While overall unemployment in AI-exposed occupations has not yet spiked, the study found one deeply troubling early signal: young workers aged 22 to 25 are already being frozen out.
Job-finding rates for this age group in AI-exposed occupations have dropped by approximately 14 to 16 percent since the launch of ChatGPT in late 2022. Entry-level hiring is slowing. The pipeline is narrowing. Graduates are emerging into a job market that is already closing doors.
These are exactly the people who, in a previous generation, would have found their social life, their identity, their daily structure through work. Without that, they need somewhere else to go. Somewhere physical. Somewhere with other people. Somewhere like your facility.
The “Great Recession for White-Collar Workers”
The researchers modelled what a genuine displacement crisis would look like. If unemployment in the most AI-exposed occupations doubled — from 3 percent to just 6 percent — it would constitute a disruption equivalent to the 2008 Great Recession, but concentrated entirely in white-collar, knowledge-economy jobs.
If the top 10 percent of exposed workers faced complete layoffs, unemployment in that group would rise from 3 percent to 43 percent. Aggregate national unemployment would jump from 4 percent to 13 percent.
These are not predictions. They are scenario models. But the gap between theoretical capability and actual adoption means the fuel for these scenarios is already loaded. What is missing is the match — and every quarterly earnings report, every cost-cutting initiative, every new AI model release brings it closer.
What This Means For The Fitness & Wellness Industry
The Bottom Line
The company that builds the AI is telling you, with data, that your industry is in the zero-exposure zone. Cooks, bartenders, lifeguards, fitness trainers — these jobs cannot be automated because they require the one thing AI fundamentally cannot provide: physical human presence and real human connection.
At the same time, millions of well-paid, well-educated workers are about to lose the one place where they had daily social contact — the office. They will need new places to go. New communities. New routines. New reasons to leave the house.
That is your market. That is your opportunity. And it is not a theory — it is Anthropic’s own data.
Employment Growth vs AI Exposure
The study found a direct, measurable relationship between AI exposure and future employment prospects: for every 10 percentage-point increase in observed AI coverage, projected employment growth drops by 0.6 percentage points.
In plain language: the more AI can do your job, the fewer of you there will be. The Bureau of Labor Statistics projections through 2034 confirm it. The occupations with highest AI coverage are the ones the government itself predicts will shrink.
And the occupations with zero coverage? They are projected to grow.
Why This Study Is Different
Previous AI impact studies — from Goldman Sachs, McKinsey, the IMF — estimated what AI might do based on task descriptions. They were theoretical. They were projections.
This study is different for three reasons:
- It uses actual usage data. Not what AI could theoretically do, but what it is doing right now, measured from real Claude interactions across hundreds of professions.
- It comes from the source. Anthropic is not observing AI from the outside. It is the company building the technology. This is the equivalent of a tobacco company publishing its own cancer research.
- It separates automation from augmentation. The study specifically weights tasks where AI replaces human work, not where it assists. This is the metric that matters for job displacement.
Read The Full Paper
Anthropic has published the complete research paper on their website. We encourage you to read it in full.
Citation: Massenkoff, M. & McCrory, P. (2026). “Labor Market Impacts of AI: A New Measure and Early Evidence.” Anthropic Research, 5 March 2026. All statistics on this page are paraphrased from the original paper with attribution. The original research is published at anthropic.com/research/labor-market-impacts.