Anthropic Research Shows What AI Is Actually Doing to Jobs Right Now

A new research report from Anthropic tracks how AI is affecting real jobs, showing early signs of change but not the job apocalypse many expected.

Every few months, a new headline appears claiming that artificial intelligence will replace millions of workers. For students entering the job market, that can feel unsettling. Should you learn AI? Avoid certain careers? Or assume everything will change soon?

A recent research report from AI giant Anthropic titled Labor Market Impacts of AI: A New Measure and Early Evidence offers one of the clearest attempts to answer that question using real data rather than speculation.

The results are more interesting and far more subtle than the usual headlines. AI is clearly beginning to affect how work is done. But the large-scale job losses many people expected have not yet appeared. Understanding why helps students think more clearly about the future of work. 

Measuring AI’s Real Impact on Jobs

One challenge in studying AI and employment is that many discussions focus on what AI could do rather than what people are actually using it for. The researchers approached the problem differently. They created a new measure called “observed exposure.” This looks at three things together:

1. Tasks that AI systems are theoretically capable of doing

2. Tasks where people are actually using AI tools

3. How much of a job consists of those tasks

To do this, they combined three major datasets: the O*NET job database, real usage data from AI systems and earlier research estimating which tasks large language models can speed up. The difference between theoretical capability and real usage is large.

For example, the report finds that AI could theoretically help with the majority of tasks in computer and mathematics jobs. But actual usage currently covers only about 33% of those tasks. AI can do far more than people are currently using it for.

That gap matters because technology adoption usually happens slowly. The internet took decades to fully reshape industries. AI may follow a similar path.
Source: Anthropic 

The Jobs Most Exposed to AI

The report also identifies occupations where AI is already being used the most. Some of the most exposed roles include:

● Computer programmers

● Customer service representatives

● Data entry workers

● Market research analysts

● Financial analysts

● Software testers

● Information security analysts

These jobs share one common feature. A large portion of the work involves text, data, or digital tasks. That happens to be exactly where current AI systems are strongest.
Source: Anthropic

By contrast, many jobs remain largely unaffected. Around 30% of workers have zero AI exposure because their tasks rarely appear in AI usage data. These include roles such as cooks, mechanics, bartenders and lifeguards.

Physical work, hands-on services and jobs requiring real-world presence remain much harder to automate. This explains an important reality about AI: it affects tasks rather than entire professions. A financial analyst might use AI to summarize reports. A marketer might use it to draft campaign ideas. But the overall job still involves judgment, coordination and decision-making. 

Low Fire-Low Hire

The most surprising result of the study is this: so far, AI has not caused measurable increases in unemployment.

The researchers compared unemployment rates for workers in highly exposed occupations with those in less exposed jobs. The trends have remained almost identical since the release of ChatGPT in late 2022. That does not mean AI has no effect either.

Instead, the early signals appear in hiring. For workers aged 22 to 25, the report finds that entry into highly exposed jobs has fallen slightly. Young workers are about 14% less likely to start jobs in those occupations compared with 2022 levels.

While companies may not be firing workers because of AI, they might be hiring fewer new ones. This pattern is common during technological change. Automation often affects new hiring before existing employment.

For students entering the workforce, the lesson is not to panic about AI. But it is also not wise to ignore it. Think about how AI actually appears in entry-level jobs.

- A marketing intern may use AI to draft social media posts.

- A research analyst might use it to summarize documents.

- A finance associate could use it to explain financial reports.

- A product manager might use it to brainstorm ideas.

In each case, AI acts more like a productivity tool than a replacement. Students who know how to use these tools effectively will likely work faster and learn quicker. Those who ignore them may struggle to keep up.

Certain small changes to a student’s routine can go a long way:

First, start treating AI like everyday software. Tools such as ChatGPT, Claude, Notion AI and Microsoft Copilot are becoming as common as Excel or PowerPoint.

Second, focus on skills that work well with AI rather than competing against it. These include problem framing, critical thinking, communication and domain knowledge.

Third, learn how to break complex tasks into smaller steps that AI tools can assist with. This is sometimes called prompting, but it is really just structured thinking.

Finally, remember that technology changes careers gradually. The internet did not eliminate jobs overnight. Instead, it created entirely new roles and industries. AI will likely do the same.

The students who benefit most will be the ones who learn to work with it early.

New research from AI giant Anthropic finds AI has not increased unemployment yet, but hiring into some AI-exposed jobs may already be slowing.

For students and early professionals, learning how to work with AI will soon be as important as learning Excel. Here is what the early evidence actually shows.

#AI #FutureOfWork #Careers #Students #ArtificialIntelligence #Productivity #Jobs #PraxisBusinessSchool