Love it or hate it, so-called generative artificial intelligence has proved its ability to make at least one type of worker more productive on the job. Using the enhanced AI boosted the productivity of customer service representatives at a Fortune 500 software firm by 14%, according to the first study to examine the emerging technology’s use by employees in a real workplace.
Notably, human support agents reported resolving more customer queries per hour when aided by a custom-built tool powered by OpenAI’s GPT technology.
Assistance from the AI also improved customer sentiment, reduced the volume of requests for managerial intervention, and even improved employee retention, presumably because it allowed service agents to have more pleasant interactions with customers, according to Erik Brynjolfsson, one of the paper’s authors and a senior fellow at the Stanford Institute for Human-Centered AI (HAI) and director of the Stanford Digital Economy Lab.
“There was less churn once they used this tool because it seemed workers were happier and enjoyed the job more,” he told CBS MoneyWatch. “We wondered if it would push them harder, but it seems to be something workers liked. Customers were happier and I’m guessing as a call center operator, it’s more enjoyable to interact with happy customers.”
Service reps interacting with customers through text chats used the custom-built AI tool to help them find answers to client questions. Specifically, the AI read interactions between customers and support agents, and generated suggested responses for the agents to use. They also had the ability to accept or reject the AI-generated text answers.
“It basically saw both what the customers and agents were saying and would give them strategically-timed hints or suggestions,” Brynjolfsson said.
For example, the AI tool would prompt reps to mention products or possible upgrades customers could make to solve their problems. It was also sensitive to tone, offering guidance to agents how to politely”communicate with clients.
The AI was trained on thousands of client-agent interactions that were labeled as either successful or unsuccessful.
“It tended to know what worked well with customers,” Brynjolfsson said.
The results: Agents generally solved problems faster, customers were happier, new employees got up to speed faster and the employer (which wasn’t identified in the study) experienced less turnover.
“People worry this is going to replace everything. I think by far the bigger effect is it augmenting us, like a calculator. It allows us to do things faster and more efficiently,” Brynjolfsson said. “I say lean in and embrace it, learn to use these tools to be more effective.”