Here’s How Much More Efficient AI Makes The Average Worker: Study

 Access to artificial intelligence rendered workers in a customer support setting 14% more efficient, according to a new working paper from Stanford University and MIT researchers.

The findings come as ChatGPT, an AI language processing tool, accrues worldwide recognition as knowledge workers leverage the system’s capabilities to execute tasks such as writing emails and fixing computer code in a matter of seconds. The academics showed that customer service employees at an unnamed Fortune 500 software company who had access to a tool based on a version of GPT answered more customer requests in the same amount of time as their colleagues who did not have access to the system.

“Access to the tool increases productivity, as measured by issues resolved per hour, by 14% on average, with the greatest impact on novice and low-skilled workers, and minimal impact on experienced and highly skilled workers,” the working paper said. “We provide suggestive evidence that the AI model disseminates the potentially tacit knowledge of more able workers and helps newer workers move down the experience curve.”

The researchers indeed found that newer and less skilled workers saw significant productivity gains, while their more experienced colleagues saw minimal improvement from the technology. The tool, which monitored customer chats and provided agents with real-time suggestions on how to respond, was “designed to augment agents,” who remained responsible for the conversation and were able to ignore the suggestions from the system.

Employees who used the AI tool saw a “decline in the time” necessary to handle an individual customer chat and an increased capacity to handle multiple chats at once, as well as a “small increase” in the portion of customer requests successfully resolved. Workers who had two months of tenure and access to the system typically performed as well as agents with six months of tenure and no access to the system.

“We posit that high-skill workers may have less to gain from AI assistance precisely because AI recommendations capture the potentially tacit knowledge embodied in their own behaviors. Rather, low-skill workers are more likely to improve by incorporating these behaviors by adhering to AI suggestions,” the researchers added. “Instead, using textual analysis, we find suggestive evidence that AI assistance leads lower-skill agents to communicate more like high-skill agents.”

The study comes as technology firms otherwise battered by recent economic tumult have seen mass market AI as a possible windfall in recent months amid drastic increase in demand for the systems among knowledge workers. Microsoft, which invested billions of dollars into ChatGPT creator OpenAI, recently announced that the system would be integrated into search engine Bing and internet browser Edge, allowing users to locate information more easily. Google revealed one day earlier that Bard, an experimental conversational AI service, would soon be added to the company’s search engine, which has dominated competitors for two decades.

Amazon, which released several mass-market AI solutions earlier this month, similarly found that coders who used AI computer programming tool CodeWhisperer completed tasks 57% faster and were 27% more likely to achieve success than those who did not use the system, which can generate real-time code suggestions.

The implications of widespread AI adoption, beyond the increase in employee efficiency, could produce considerable technological unemployment: Goldman Sachs noted in a recent report that mass market AI systems could have “potentially large macroeconomic effects” and produce “significant disruption” in worldwide labor markets, affecting as many as 7% of positions in the United States, especially in sectors reliant upon office work.