One of the largest real-world studies of AI at work looked at customer support, and the results are hard to ignore for any business fielding the same questions all day.
Most debates about AI are theoretical. This one wasn't. Economists measured what actually happened when thousands of real support agents were given an AI assistant, and tracked their output for months.
The study
In "Generative AI at Work" (NBER Working Paper 31161, 2023), economists Erik Brynjolfsson, Danielle Li and Lindsey Raymond studied an AI assistant used by 5,179 customer-support agents at one company.
On average, agents resolved 14% more issues per hour. But the effect wasn't evenly spread: newer and lower-skilled agents improved by 34%, while the most experienced agents saw little change. In effect, the AI spread the know-how of the best agents to everyone else, and customers reported more positive interactions too.
Why it works
Support teams often see recurring questions about hours, order status, refunds, availability and product details. An AI assistant can help agents find approved answers or draft replies, but the right level of automation depends on the topic, the quality of source information and the cost of an incorrect answer.
What the study does—and does not—show
The study measured an assistant supporting human agents at one company. It did not test an autonomous support service or establish results for other businesses. Launchzy's AI Customer Support is a separate workflow: its channels, approved answers and human handoffs should be scoped and tested against the business's actual support process.
The result is evidence about that deployment—not a forecast for every team and not a guarantee that a different tool will achieve the same outcome.
Give customers instant answers, 24/7
See how an AI Support agent would handle your repetitive tickets and free up your team.
Book your free audit callSource
- Brynjolfsson, Li & Raymond, "Generative AI at Work," NBER Working Paper 31161 (2023).