Anthropic launches tool to monitor jobs lost to AI systems

Anthropic PBC today announced it has introduced a detection tool that measures artificial intelligence’s displacement of jobs.

The forecasts so far of what AI will do to the labor force have ranged from dystopian to mild. No one knows what will happen, although Anthropic PBC Chief Executive Dario Amodei has been at the forefront of raising the alarm concerning what might happen.

In 2025, he warned that all entry-level white-collar jobs could be gone within the next five years, causing a spike in U.S. unemployment rates. He repeated that in a recent essay when he asked if the world was prepared for more advanced AI systems. He also said AI doomerism isn’t productive or rational, but a head-in-the-sand approach is not useful, either.

In the paper released today, Anthropic economists Maxim Massenkoff and Peter McCrory acknowledged that we are still struggling to understand what will happen in the marketplace, with some forecasts often contradicting others. The “impact of major economic disruptions on the labor market is often unclear,” they explained.

Their tool combines a range of measurements. The scientists studied what is theoretically possible with AI and what tasks could be replaced with those systems. It then looks at which tasks are currently being performed by the same large language models.

There are no surprises with what they believe is theoretically possible – office and admin work, management, business and finance, computer and math, with legal, art and media, and architecture and engineering not far behind.

The most exposed occupation to disruption, based on LLM use from the Anthropic Economic Index, is computer programming, with a 75% coverage. This was followed by Customer Service Representatives (70.1%) and Data Entry Keyers (67.1%). At the bottom, occupations where Anthropic sees 0% coverage with AI were cooks, motorcycle mechanics, lifeguards, bartenders, dishwashers and dressing room attendants.

Importantly, the scientists said unemployment rates in the most exposed occupations have not increased at higher rates than those occupations that have seen no AI exposure. “Using survey data from the U.S., we find no impact on unemployment rates for workers in the most exposed occupations, although there’s tentative evidence that hiring into those professions has slowed slightly for workers aged 22-25,” they explained.

It’s these measured, analytical steps to studying AI job disruption that are needed, they wrote. When we see the changes take place in real time, we can make adjustments. “An established approach may help future observers separate signal from noise,” they said.

Photo: Unsplash