Generative AI at Work
Erik Brynjolfsson, Danielle Li, Lindsey Raymond
The Quarterly Journal of Economics
The research approach
Examines a staggered deployment of a generative AI assistant among customer-support agents at a business software firm, comparing outcomes as access becomes available.
What the research finds
AI assistance increased resolved issues per hour on average, with larger gains among less experienced workers; effects varied substantially across workers.
Why it matters for management
Provides evidence for evaluating augmentation, workforce development and the distribution of benefits when organizations introduce AI assistance.
Scope & limitations
Evidence comes from one firm's customer-support setting and a particular tool; results should not be assumed to apply equally to other occupations or technologies.



