Is AI Really Taking All the Jobs? Yale’s Data Says Not Yet.

Earn CPE with this course

I see the headlines every day. Another tech company announces layoffs. The reason is always AI. It feels like a massive shift is happening right now. A robot revolution wiping out jobs left and right.

But is that what’s really going on?

I wanted to check the hype against the data. So I dug into a recent report from the Yale Budget Lab. They analyzed US labor data from 2022 through mid-2025 to see if they could find AI’s fingerprints on the job market.

Is the Job Mix Changing Faster Than Ever?

First the report asks a simple question. Is the AI era changing jobs faster than past tech waves like the PC or the internet?

To figure this out they used a “dissimilarity index.” It sounds complicated but it’s not. It’s just a score that measures how much the mix of jobs has changed over time. A score of 7 means 7% of workers would have to switch occupations to get back to the job mix we started with.

The finding? The job mix is changing a bit faster now than during the internet rollout. But the trend started back in 2021 before ChatGPT was a household name. The change is modest and not historically unusual. We are not seeing a massive unprecedented disruption.

Are “AI-Exposed” Jobs Disappearing?

This is the real question. Are the specific jobs AI is supposed to be good at the ones seeing losses? The researchers looked at this in two ways.

1. The Theory (OpenAI’s Data)

First they used an “exposure” metric from OpenAI. This is a theoretical risk score. It’s based on how much a model like GPT-4 could reduce the time it takes to do tasks in a given job. They grouped jobs into low medium and high exposure.

The result? Flat. The share of workers in high-exposure jobs has been stable since ChatGPT’s launch. Their unemployment rate hasn’t spiked. The jobs AI is supposed to replace are not disappearing.

2. The Reality (Anthropic’s Data)

Next they looked at observed usage data from Anthropic’s AI model Claude. This isn’t theory. This is how people are actually using AI. They split the usage into two types:

• Augmentation: AI helps a person do their work. The human is in the driver’s seat.
• Automation: AI does most of the work. The human oversees it.

The result? Also flat. Unemployment is not clustering in jobs with high automation or augmentation. There is no clear sign that places where people are actively using AI are seeing any notable change in employment.

But The Data Has Its Limits

The Yale researchers are clear. Their metrics are just proxies. They don’t have the full picture.

• Exposure is just a theory. Two jobs with the same risk score could have very different real-world AI adoption. A software developer and a clerical worker might adopt AI at completely different speeds.
• Usage data is skewed. The Anthropic data only reflects Claude users. That user base is heavy on coders and writers. It doesn’t capture the massive user bases of ChatGPT Gemini or Copilot.
• The data is partial. The report can’t see everything. It misses a lot of data from inside companies and from API usage.

This doesn’t invalidate the findings. It just means we’re seeing a small slice of reality. But based on this slice there is no clear sign that AI is driving unemployment.

The Real Disruption Takes Decades Not Months

The report’s final point puts everything in perspective. Large-scale labor market effects from big technologies like the PC the internet and now AI take decades to play out. Not months.

The biggest job shifts in US history happened in the 1940s and 50s due to industrial changes. Not during the PC and internet booms. The current rate of change is nowhere near that.

AI is happening a little faster than the internet rollout but not by much. It is likely to take years or decades to see the big organizational changes everyone is predicting.

Key Takeaways

After reading through the Yale study here’s where I landed.

• No Clear Link: The data shows no visible link between AI exposure and changes in employment or unemployment levels so far.
• Not Unprecedented: The current rate of job market change is modest. It isn’t historically unusual and the trend began before the generative AI boom.
• Hype vs. Reality: The media narrative of massive AI-driven job loss is not supported by the current labor data.
• It’s a Marathon: The real impact of AI will likely play out over decades. The immediate focus should be on tracking how tasks are changing not predicting mass job loss.

Want to earn CPE for this topic?

• Compare Options: See how we stack up against others in our 2025 Flexible CPE Guide
• Understand the Format: Read how Nano-Learning works for CPAs.
• Check Your State: Ensure you are compliant with our State Requirements Guide.

Further Reading

Here are some links to the original report and other articles that cover these findings:

• The Original Report (Yale Budget Lab): Evaluating the Impact of AI on the Labor Market: Current State of Affairs
• Forbes: More AI, Fewer Jobs? No Sign Yet Of An Apocalypse, Says Yale Study
• Built In: The Job Market Might Be a Mess, But Don’t Blame AI Just Yet
• Workplace Insight: Yale study finds little evidence that AI is taking people’s jobs