{"slug": "anthropic-releases-economic-research-on-claude-code-usage-reveals-humans-still", "title": "Anthropic releases economic research on Claude Code usage, reveals humans still do most of the thinking", "summary": "Anthropic released an economic study of 400,000 Claude Code sessions, finding humans handle 70% of planning decisions while AI executes 80% of tasks. Domain experts generate over twice as many actions per prompt as novices, and debugging sessions dropped by nearly 50% as users shifted to deployment and data analysis. The report indicates AI coding tools are expanding beyond software engineering into other occupations.", "body_md": "# Anthropic releases economic research on Claude Code usage, reveals humans still do most of the thinking\n\nA study of 400,000 coding sessions shows AI handles execution while humans drive strategy, and domain experts get dramatically more out of the tool than novices.\n\nAnthropic just dropped a research report that reads like a performance review for its AI coding agent. The verdict: Claude Code is great at doing things, but humans are still the ones deciding what to do.\n\nThe report, titled “Agentic coding and persistent returns to expertise,” analyzed roughly 400,000 Claude Code sessions from about 235,000 users between October 2025 and April 2026. It’s part of Anthropic’s broader Economic Index series, which has been quietly building one of the more rigorous datasets on how people actually use AI tools in the wild.\n\n## The division of labor is stark\n\nUsers are responsible for approximately 70% of planning decisions, essentially telling Claude what needs to happen. Claude, in turn, handles about 80% of execution decisions, figuring out how to get it done.\n\nDomain expertise, the kind of knowledge that goes beyond just knowing how to code, makes a massive difference. Expert users coax Claude into generating over twice as many actions per prompt and five times more output compared to novices.\n\n## Debugging is out, deployment is in\n\nOne of the more striking trends in the seven-month dataset is a behavioral shift in what people actually use Claude Code for. Debugging sessions dropped by nearly 50%. Where did that time go? Toward what Anthropic calls “end-to-end agentic workflows,” things like deploying code, running applications, and performing data analysis.\n\nThe practical value of tasks also climbed. Anthropic estimates that typical task value rose by about 25% over the study period, benchmarked against prevailing freelance job rates.\n\nThe March 2026 Economic Index edition noted that coding tasks were migrating away from Claude.ai’s consumer chat interface toward the more powerful, API-driven Claude Code environment. That earlier report also documented that coding represented roughly 35% of all Claude.ai conversations, making it the single largest category. The new data shows coding tasks are spreading beyond traditional software engineering into other occupations, with non-software roles as the fastest-growing user group in the Claude Code dataset.\n\n## What this means for investors\n\nThe finding that success rates between intermediate and expert users show only modest gaps is worth flagging. It means the barrier to getting functional output from AI coding tools is lower than you might expect. The expertise premium shows up in volume and sophistication of output, not in whether the code runs at all.\n\nThe risk here is overextrapolation. This is a single company’s dataset from its own product, covering a specific seven-month window. The sample is large, roughly 400,000 sessions, but it represents Claude Code users specifically, not the entire AI coding market.\n\n**Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our\n\n[Editorial Policy](https://cryptobriefing.com/editorial-policy/).", "url": "https://wpnews.pro/news/anthropic-releases-economic-research-on-claude-code-usage-reveals-humans-still", "canonical_source": "https://cryptobriefing.com/anthropic-claude-code-economic-research/", "published_at": "2026-06-16 21:13:39+00:00", "updated_at": "2026-06-16 21:20:16.056160+00:00", "lang": "en", "topics": ["ai-agents", "ai-research", "ai-tools", "large-language-models", "developer-tools"], "entities": ["Anthropic", "Claude Code", "Claude.ai"], "alternates": {"html": "https://wpnews.pro/news/anthropic-releases-economic-research-on-claude-code-usage-reveals-humans-still", "markdown": "https://wpnews.pro/news/anthropic-releases-economic-research-on-claude-code-usage-reveals-humans-still.md", "text": "https://wpnews.pro/news/anthropic-releases-economic-research-on-claude-code-usage-reveals-humans-still.txt", "jsonld": "https://wpnews.pro/news/anthropic-releases-economic-research-on-claude-code-usage-reveals-humans-still.jsonld"}}