Generative AI at Work
Summary¶
Brynjolfsson, Li, and Raymond study the staggered rollout of a generative-AI-based conversational assistant across 5,172 customer-support agents and find that access to AI assistance raises worker productivity, measured as issues resolved per hour, by 15% on average. The gains are highly heterogeneous: less experienced and lower-skilled workers improve in both speed and quality, while the most experienced and highest-skilled workers see small speed gains and small declines in quality. The tool also appears to facilitate worker learning, improve English fluency among international agents, and improve the experience of work (customers are more polite and less likely to escalate to a manager).
Contribution¶
One of the first large-scale, in-the-wild quasi-experimental estimates of generative-AI productivity effects in a real workplace, with a clear identification strategy (staggered adoption), and the most-cited evidence to date that GenAI can disseminate the behaviours of the most productive workers to less experienced peers.
Method¶
Quasi-experimental analysis of the staggered introduction of a generative-AI conversational assistant; N=5,172 customer-service agents; outcome is issues resolved per hour, with heterogeneity analysis across worker tenure and skill and additional outcomes on learning, language, and customer behaviour.
Relevance to RISE¶
A core empirical anchor of the RISE research-productivity thread and the productivity-cluster alongside noy2023experimental and filimonovic2025genai. The skill-compressing pattern (largest gains to low-skill workers) and the "learning" channel (AI assistance accelerates skill acquisition) are directly relevant to how human-ai-research-collaboration should be designed in RISE catalog projects, and the sociotechnical observations about customer behaviour speak to second-order effects of deploying GenAI in knowledge work.
Critique / open questions¶
The setting is customer support, not knowledge production, and "issues resolved per hour" is a narrow productivity measure. The small quality decline for top-skill workers raises unresolved questions about how GenAI should be deployed when the workforce is highly heterogeneous.
Key quotes¶
"We study the staggered introduction of a generative AI–based conversational assistant using data from 5,172 customer-support agents. Access to AI assistance increases worker productivity, as measured by issues resolved per hour, by 15% on average, with substantial heterogeneity across workers."
"Less experienced and lower-skilled workers improve both the speed and quality of their output, while the most experienced and highest-skilled workers see small gains in speed and small declines in quality."
"We also find evidence that AI assistance facilitates worker learning and improves English fluency, particularly among international agents."