Economic Index report: Cadences
A · Strong evidence Anthropic's economics team (Peter McCrory, Maxim Massenkoff, Zoe Hitzig and others) combines privacy-preserving usage telemetry with a linked survey of about 9,700 users, a strong dataset that comes entirely from the vendor's own product.
Summary based on the full text·Original: about 35 minutes to read
In short
The sixth Anthropic Economic Index report, published June 26, 2026 by Anthropic's economic research team. It adds daily and hourly sampling, a classifier for the output ("artifact") of each conversation, and the first results of the Anthropic Economic Index Survey, linked to usage for about 9,700 respondents. Usage follows the work week and daily routines, and compute rises with the wage level of the work. People who delegate the most expect AI to take on more of their tasks and are also the most optimistic about their pay and job security.
Key numbers
- 0.37 pointsaverage AI-autonomy gap (1-5 scale) between Claude Code and chat/Cowork for the same output types
- 13 vs 1median human prompts to produce a blog post in chat/Cowork vs Claude Code
- 10%linked survey respondents rating loss of their own job in the next year as likely or very likely
- 2.07xtokens consumed by a top-tercile-wage occupation conversation vs a bottom-tercile one
As reported by the source.
English·Web page·opens anthropic.com
What it reports
- Personal-use share of chat and Cowork conversations moves from about 35% on weekdays to just under 50% on weekends.
- 93% of conversations produce an identifiable artifact: explanations 17%, documents and reports 15%, guidance 11%. Work conversations most often produce documents and reports (20%).
- Top-wage-tercile conversations consume 2.07 times the tokens of bottom-tercile ones; about 44% of the gradient comes from output mix. Claude produces 1.34 times more output per turn and users take 1.53 times as many turns.
- Autonomy is higher in Claude Code than chat/Cowork for 26 of 31 output types; the average gap is 0.37 points on a 1-5 scale and 0.26 points when both use Sonnet. 54% of Claude Code sessions run on Opus against 10% of chat/Cowork.
- Survey: close to 6 in 10 expect a higher share of their tasks to be doable by AI in 12 months; over a third expect AI to do most or nearly all of their work tasks next year.
- 10% rate losing their own job in the next year as likely or very likely; of those, 38% attribute the forecast to AI. Over a third put the chance of a junior colleague losing their job above 60%.
- Workers with 15+ years of experience report about 10 percentage points lower AI task share than first-year workers.
- Self-reported gains: speed 86%, scope 82%, quality 69%; 68% report learning more, 57% say their skills became more valuable.
- Women are 12% of the linked sample; controlling for occupation, their automation share is 7.3 percentage points lower and Claude Code share 6.3 points lower.
How they got it
Continuous privacy-preserving sampling of Claude conversations, classified by task, artifact, autonomy, reading level and work/personal use, with tasks mapped to US occupations and BLS wages. The survey, launched April 2026, was linked to up to 20 sampled sessions per respondent from mid-May to early June; respondents with fewer than five sessions were excluded.
Limits
- The survey is not representative: computer and mathematical occupations are about 30% of respondents versus 4% of US employment, management 23% versus 7%.
- Optimism among heavy delegators may reflect selection: enthusiasts delegate more. The authors cannot rule this out.
- Learning, productivity and skill value are self-assessed; Anthropic is reporting on its own product.
Why it matters for companies
The same model produces more delegation in an agentic tool than in chat, so the product surface a company deploys shapes how work changes, not only the model it licenses. Experienced staff consistently say judgment, context and relationships are what AI lacks, which points to where reskilling and role design should focus. Early-career anxiety is highest, which matters for graduate hiring and retention. The finding that compute scales with the value of work is a direct input for budgeting AI consumption by function.
Summary based on the full text. Read the original for the full argument and data.Read the original on anthropic.com
- Tags
- AI adoption