SecondBrain
Ask the Brain
Index/Entityupdated Sat Jun 27 2026 08:00:00 GMT+0800 (Philippine Standard Time)

Anthropic Economic Index

anthropicresearcheconomic-indexai-and-workautomationsurvey

Anthropic Economic Index

Anthropic's in-house research program on the economic impact of Claude usage. Multiple reports per year, plus the linked Economic Index Survey (launched April 2026). The reports are the closest first-party empirical data on what people actually use frontier LLMs for, which makes them the de facto benchmark for the next-tier "AI and labor" question this vault tracks.

Cadence so far

  • March 2026 — the empirical observation that Steven Brovich cites in A Leaders Guide to Advanced Team Structures (AWS Events): AI's observed exposure on real jobs is far below the theoretical exposure that LLM-could-do-this exercises estimate. Hiring of younger talent in exposed occupations slowed ~14% since ChatGPT (Nov 2022). The most-exposed workers (computer programmers, customer service, data entry) are disproportionately older, more educated, and better paid than average — the opposite of the "AI replaces the lowest-skilled first" narrative.
  • June 2026 — "Cadences" (Anthropic Economic Index Cadences Report (June 2026)) — higher-rate sampling for hourly resolution; new artifact classifier; data broken out for Claude chat + Cowork / Claude Code / 1P API; first results from the EI Survey.

What's new in the methodology

  • Sample rate raised so daily and hourly patterns are visible (prior reports used 7-day samples)
  • New artifact classifier labels the >30-category output of each conversation — see Artifacts (Claude Output)
  • Three-cohort split: chat + Cowork ("Claude conversations") / Claude Code / 1P API, aggregated monthly
  • Anthropic Economic Index Survey (launched April 2026): self-reported expectations + experiences, linked to actual usage via Anthropic's privacy-preserving system

The EI Survey

The Chapter 3 component of the June 2026 report. Combines:

  • A self-reported survey (launched April 2026) on how people perceive AI to be changing their work and what they want from AI
  • Linkage to actual Claude-usage data via privacy-preserving infrastructure
  • Cluster analysis of users by how they use Claude (more vs less automated workflows)

Headline finding (per the report preview): "People who use Claude in the most automated way expect AI to take on more of their tasks in the next year, yet feel the most optimistic about what that means for their work, anticipating positive impacts on pay, job security, and meaning."

This is a counter-intuitive correlation worth holding onto carefully — it's compatible with both a selection story (optimists adopt more aggressively) and an experience story (delegation reveals delegation is good for me). The captured source body is truncated before the mechanism is unpacked.

Why this is its own entity page (and not just a section on Anthropic)

  • Recurring program — multiple reports per year, will keep producing the cleanest "AI and work" empirical anchors
  • Cited externally — the March report is already the data anchor for Steven Brovich's AWS Events keynote on agentic-world team structures; expect cross-references to compound
  • Methodology evolves issue-to-issue — worth its own page so changes (e.g. adding the artifact classifier in June 2026) stay legible

Cross-references

  • Tasks to Responsibilities Shift — Cowork-separated-from-chat is the consumer-side evidence of the shift
  • Code Is Free — the EI's compute-vs-wage scaling is the missing empirical anchor under Lopopolo's thesis
  • FOBO (Fear of Becoming Obsolete) — the automated-use → optimism survey result is the empirical counter to the dominant FOBO narrative
  • Hourglass Organization — the March 2026 EI finding that hiring of younger talent slowed ~14% is the entry-rung data point under Brovich's "protect the juniors" prescription

Sources