Escape Meridian

· Feb 10, 2025

Anthropic introduces an index of AI use across work

Markdown · JSON · Original source

Published February 10, 2025, introduces the Anthropic Economic Index to track how AI is used in work and the economy. Its first paper analyzes about one million Free and Pro conversations on .ai, using Clio, an automated analysis tool, to match conversations to occupational tasks. Anthropic reports that use was concentrated in software development and technical writing; about 36% of occupations had AI use in at least a quarter of their tasks, while about 4% used AI across at least three-quarters. It reports more augmentation—AI working with a user—than automation, where AI performs tasks directly. Anthropic says the data did not show jobs being entirely automated, but cautions that conversations may not have been work-related and that it cannot know how users used Claude’s responses. The analysis covers Claude.ai Free and Pro, not , Team, or Enterprise users, and Anthropic says its uses should not be treated as representative of AI use in general. Anthropic plans repeated analyses and data releases to track changes, but says its research does not provide policy prescriptions; preparing for labor-market effects requires evidence, values, and experience from broad perspectives.

Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.

AnthropicArticle · anthropic.comOpen original source

Original source

· Article

Source details

Original source date: Feb 10, 2025 (2025-02-10T00:00:00.000Z).

Date basis: Date stated on the publisher’s page.

Added to Escape Meridian: 2026-11-15T11:18:24.508Z. First observed: 2026-10-18T09:19:19.996Z. Retrieved: 2026-10-30T01:39:45.401Z.

Referenced by

Recorded citations and public file attachments. A reference alone does not establish agreement or influence.

No public references recorded here yet.

Reference data (JSON)

Limits

Text as retrieved on October 30, 2026; originally published February 10, 2025. The analysis covers .ai Free and Pro conversations, not , Team, or Enterprise users; says the dataset is not a representative sample of AI use in general. Anthropic says it cannot know whether a conversation was for work or how users used Claude’s responses; it also notes that Clio may have classified some conversations incorrectly.

Open question

In ’s planned repeated analyses, will AI use remain concentrated in selected tasks, or spread across a larger share of occupations’ tasks?

Source-stated

Anthropic says its first paper uses about one million Free and Pro Claude.ai conversations, organized by occupational task with Clio.

We used Clio on a dataset of approximately one million conversations with Claude (specifically, Free and Pro conversations on Claude.ai)

Source-stated

Anthropic reports that AI use appeared in at least a quarter of the tasks associated with roughly 36% of occupations, and across at least three-quarters of tasks in approximately 4%.

Over one-third of occupations (roughly 36%) see AI use in at least a quarter of their associated tasks, while approximately 4% of occupations use it across three-quarters of their associated tasks.

Source-stated

Anthropic reports a greater share of augmentation than automation in its analysis.

Overall, we saw a slight lean towards augmentation, with 57% of tasks being augmented and 43% of tasks being automated.

Source-stated

Anthropic says its dataset did not show jobs being entirely automated, and cautions that it does not represent AI use in general.

As we predicted, there wasn’t evidence in this dataset of jobs being entirely automated

Source-stated

Anthropic says its research does not provide policy prescriptions and calls for broad perspectives on preparation for AI’s labor-market impact.

Our research gives data on how AI is being used, but it doesn’t provide policy prescriptions.

Source and contributor context

Anthropic · Publisher

The publisher is identified here. No individual author has been inferred.

Source media

Media retrieval status: unchecked.

Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.

API

An application programming interface: a defined way for software to request data or actions from another system.

Anthropic

Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.

Claude

Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.

Editorial provenance

Version 1. Publication ID: e600f33e-ab4d-b4b6-e0bd-b1fcf37d677a.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-15T11:13:48.134Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge_admitted_by_model.

Observation: ac7f6412ade8170463aa54e326eb50e7685fc20f221fd5daf5373bd255abd9c0. Source text hash: 707714adcb78bc821f99a0aaf80beed4892f07273c7d9649f51469ebecd1d48f.

Revision and source-check history

Version 1 · 2026-11-15T11:18:24.508Z

Anthropic introduces an index of AI use across work

Published February 10, 2025, introduces the Anthropic Economic Index to track how AI is used in work and the economy. Its first paper analyzes about one million Free and Pro conversations on .ai, using Clio, an automated analysis tool, to match conversations to occupational tasks. Anthropic reports that use was concentrated in software development and technical writing; about 36% of occupations had AI use in at least a quarter of their tasks, while about 4% used AI across at least three-quarters. It reports more augmentation—AI working with a user—than automation, where AI performs tasks directly. Anthropic says the data did not show jobs being entirely automated, but cautions that conversations may not have been work-related and that it cannot know how users used Claude’s responses. The analysis covers Claude.ai Free and Pro, not , Team, or Enterprise users, and Anthropic says its uses should not be treated as representative of AI use in general. Anthropic plans repeated analyses and data releases to track changes, but says its research does not provide policy prescriptions; preparing for labor-market effects requires evidence, values, and experience from broad perspectives.

Original source

· Article

Source details

Original source date: Feb 10, 2025 (2025-02-10T00:00:00.000Z).

Date basis: Date stated on the publisher’s page.

Added to Escape Meridian: 2026-11-15T11:18:24.508Z. First observed: 2026-10-18T09:19:19.996Z. Retrieved: 2026-10-30T01:39:45.401Z.

Limits

Text as retrieved on October 30, 2026; originally published February 10, 2025. The analysis covers .ai Free and Pro conversations, not , Team, or Enterprise users; says the dataset is not a representative sample of AI use in general. Anthropic says it cannot know whether a conversation was for work or how users used Claude’s responses; it also notes that Clio may have classified some conversations incorrectly.

Open question

In ’s planned repeated analyses, will AI use remain concentrated in selected tasks, or spread across a larger share of occupations’ tasks?

Source-stated

Anthropic says its first paper uses about one million Free and Pro Claude.ai conversations, organized by occupational task with Clio.

We used Clio on a dataset of approximately one million conversations with Claude (specifically, Free and Pro conversations on Claude.ai)

Source-stated

Anthropic reports that AI use appeared in at least a quarter of the tasks associated with roughly 36% of occupations, and across at least three-quarters of tasks in approximately 4%.

Over one-third of occupations (roughly 36%) see AI use in at least a quarter of their associated tasks, while approximately 4% of occupations use it across three-quarters of their associated tasks.

Source-stated

Anthropic reports a greater share of augmentation than automation in its analysis.

Overall, we saw a slight lean towards augmentation, with 57% of tasks being augmented and 43% of tasks being automated.

Source-stated

Anthropic says its dataset did not show jobs being entirely automated, and cautions that it does not represent AI use in general.

As we predicted, there wasn’t evidence in this dataset of jobs being entirely automated

Source-stated

Anthropic says its research does not provide policy prescriptions and calls for broad perspectives on preparation for AI’s labor-market impact.

Our research gives data on how AI is being used, but it doesn’t provide policy prescriptions.

Source and contributor context

Anthropic · Publisher

The publisher is identified here. No individual author has been inferred.

Source media

Media retrieval status: unchecked.

Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.

API

An application programming interface: a defined way for software to request data or actions from another system.

Anthropic

Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.

Claude

Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.

Editorial provenance

Version 1. Publication ID: e600f33e-ab4d-b4b6-e0bd-b1fcf37d677a.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-15T11:13:48.134Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge_admitted_by_model.

Observation: ac7f6412ade8170463aa54e326eb50e7685fc20f221fd5daf5373bd255abd9c0. Source text hash: 707714adcb78bc821f99a0aaf80beed4892f07273c7d9649f51469ebecd1d48f.

passed · 2026-11-15T11:13:48.134Z · anthropic/claude-opus-5.5

Publication: e600f33e-ab4d-b4b6-e0bd-b1fcf37d677a

attribution — pass: The reading consistently attributes the findings, caveats and plans to Anthropic, which is the publisher and author of this single announcement. The evidence is one piece with its own argument and findings, not a listing of other pieces.

we're launching the Anthropic Economic Index, an initiative aimed at understanding AI's effects on labor markets and the economy over time

summary — pass: The summary leads with what Anthropic launched and explains its method. It explains Clio, augmentation and automation in plain language, and API is defined on the page. It also conveys the source's main numbers, caveats and its stance that it offers no policy prescriptions. The question follows from the source's own plan to monitor depth of AI use within occupations. The title is accurate.

we’ll be able to monitor changes in the depth of AI use within occupations

date — pass: The displayed date of February 10, 2025 matches the UTC day of evidence.originalPublished. Although the page was modified on 2026-10-23, the retrieved text reads as the original launch announcement, and the reading relies on no passage describing later events. Forecasts and plans, such as repeated analyses, are reported as Anthropic's plans. The page is a publisher page, so no archive disclosure is needed.

we’ll repeat many of the analyses above over time to help track the societal and economic changes that are likely to occur

support — pass: Each claim matches the source text: about one million Free and Pro conversations, the 36% and 4% figures, the 57/43 augmentation lean, no jobs entirely automated, the non-representativeness caveat, and no policy prescriptions. The cited passages are core body text of the launch article and do not look like later additions. The reading implies no benefits or costs beyond the source.

we don’t argue that the uses in our dataset are a representative sample of AI use in general

media — pass: No media candidates were supplied, and the reading selects none. Nothing was invented.

Challenge or correct this reading

Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.