Anthropic Predicts Which Jobs AI Will Hit First

Anthropic Predicts Which Jobs AI Will Hit First

by Paul Roetzer and Mike Kaput

Humanities, Social Sciences & Thought Economics & Finance KCEconomicsKCFLabour
🎙 Paul Roetzer and Mike Kaput 👥 31K 📅 March 13, 2026 ⏱ 16 min 👁 4K 📄 news review 🧭 2026-08-16
Available in: English (current) Français

Keywords

observed exposureknowledge workerswhite-collar jobsunemploymentsocial contract

Summary

In this episode of The Artificial Intelligence Show, hosts Paul Roetzer and Mike Kaput discuss a recent Anthropic study on AI’s impact on white-collar jobs. They introduce the concept of ‘observed exposure,’ which compares theoretical AI capabilities with actual usage data from Claude. The study finds that while AI could theoretically handle 94% of knowledge worker tasks, it currently covers only 33%. The most exposed workers are highly educated and well-paid, with computer programmers, customer service reps, and data entry keyers at the top. The hosts note that despite no systematic unemployment increase yet, there are early warning signs for Gen Z, with a 14% drop in job-finding rates for young workers in exposed fields. They emphasize the gap between theory and reality, cautioning against a false sense of security. The discussion extends to the broader societal implications, including the social contract between workers, employers, and society, and potential policy responses like universal basic income or an AI tax. The hosts stress the need for proactive conversations and policies to address the inevitable changes.

174 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides valuable insights into Anthropic’s research, offering specific statistics and a clear explanation of the observed exposure metric. The hosts effectively argue that the gap between theoretical and actual AI adoption is significant and may lead to a false sense of security. They support their points with references to prior studies and personal experiences, but the argumentation sometimes veers into speculative territory, such as discussing an AI tax or universal basic income without deep analysis. The discussion is well-structured, moving from data to implications, but lacks rigorous counterarguments or alternative perspectives.

Scientific Rigor, Source Quality, Title Accuracy

The hosts accurately reference the Anthropic study and mention related work like GDP-eval and the OpenAI/Microsoft paper. They also point to O*NET as a source for task-level analysis. However, they do not provide direct links to the study or these sources in the description, relying instead on general references. The title accurately reflects the content, focusing on Anthropic’s predictions. The discussion is scientifically grounded but includes personal opinions and hypothetical scenarios, which slightly reduces overall rigor.

184 words

Title / Content Match

The title accurately reflects the content, which focuses on Anthropic's predictions about AI's impact on jobs.

Quality & Reliability

7/10

The hosts accurately summarize Anthropic's research on observed exposure, citing specific statistics and contextualizing with prior studies. They clearly distinguish between theoretical and observed exposure, and acknowledge limitations of the data. However, they introduce speculative elements (e.g., AI tax, social contract) without rigorous analysis, and the discussion is framed for a business audience.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • Anthropic's study on observed exposure — Main study discussed, measuring AI's actual impact on white-collar jobs.
  • GDP-eval — Previous research by Anthropic on economically valuable tasks.
  • O*NET database — Government database breaking down occupations into tasks.
  • OpenAI/Microsoft paper on AI and jobs — 2023 paper analyzing task-level exposure using O*NET.
  • Jobs GPT — Custom GPT by Paul Roetzer for task-level job exposure analysis.
  • Andrew Yang interview — Podcast interview discussing universal basic income and social contract.

Concurring Sources

  • Anthropic's study on observed exposure — Main study discussed, measuring AI's actual impact on white-collar jobs.
  • GDP-eval — Previous research by Anthropic on economically valuable tasks.
  • OpenAI/Microsoft paper on AI and jobs — 2023 paper analyzing task-level exposure using O*NET.

Dissenting Sources

  • Potential counterarguments on AI job displacement — The video does not present any discordant sources, but some economists argue that AI may create more jobs than it displaces.

External References

Contribution & Novelties

The video provides a clear and accessible breakdown of Anthropic’s observed exposure metric, highlighting the gap between theoretical and actual AI adoption. It adds value by connecting the research to broader societal questions about the social contract and policy responses. The hosts offer practical advice for companies to establish their own evals. However, the discussion is largely a summary of existing research, with limited original analysis.

Pour aller plus loin :

101 words

Radar Profile

The radar profile shows moderate scores across all dimensions, with slightly higher scores in information quantity and quality, reflecting a well-summarized discussion of research. The technical level is moderate, suitable for a business audience, and the overall reliability is good but not exceptional due to speculative elements.

Reliability 7/10

💬 Sur les 0 commentaires analysés, aucune tendance n'est observable.