What happened
Anthropic published on 16 June 2026 a quantitative economic research paper analysing ~400,000 Claude Code sessions from ~235,000 users between October 2025 and April 2026. The headline finding: 'the value of the typical task, which we estimate through a comparison to freelance job postings, rose in almost every kind of work—about 25% on average' over the seven-month observation window. The study introduces a privacy-preserving framework that classifies agentic sessions by work mode, human-AI division of labour, and success rate. A key structural finding is that domain expertise—not coding proficiency—is the primary predictor of successful task completion: non-engineers succeed at nearly the same rate as software engineers when using Claude Code. Over the period, debugging sessions fell by nearly half and usage shifted toward end-to-end agentic deployment. The paper also observes that Claude is now handling increasingly complex, higher-value tasks autonomously.
Why it matters
This is the most empirically grounded study yet of how agentic AI is reshaping knowledge work at scale; its finding that domain expertise—not coding skill—drives AI productivity gains has direct implications for workforce strategy, hiring, and talent development decisions that boards and CHROs need to make now.
Action needed
Share with CHRO and workforce strategy leads: the 'domain expertise > coding proficiency' finding should inform AI upskilling programmes, role redesign, and how the organisation measures AI-enabled productivity gains.