Generative AI: Friend or Foe in the Call Center?
In this increasingly digital world, sometimes the only thing that will satisfy our customer service needs and make us feel we’ve been understood is live interaction with a fellow human being. Yet many companies remain challenged by staffing shortages as today’s expectations of operational efficiency and consumer experience test the endurance of agents facing back-to-back calls.
Enter ChatGPT, a model of AI that generates human-like responses for almost any question or request. Unlike its predecessor, machine learning that uses algorithms to crunch data and make informed predictions, generative AI not only recognizes patterns but can also generate fresh content from the data it was trained on.
Hence, AI “chatbots,” when used in a customer service setting, are able to speak to the customer as a person would, with tone and inflection. And because they are trained to sense heightened emotion, such as frustration or impatience, they can be tasked with determining whether a call should be forwarded to a live service agent.
Recently, some contact centers have begun using conversation analytics programs to scan and analyze conversations between agents and customers. Under this scenario, AI-powered chatbots identify words and sentiments customers are expressing to find patterns, which they can then quickly interpret to recommend what an agent should say and do next.
In a recent Wall Street Journal interview, call center agents across a range of companies said although they value AI’s ability to help them make decisions quickly, they would object to being forced to follow AI-generated scripts or coaching. This is consistent with previous research showing that workers subjected to close monitoring by algorithms or who have little control over how they work are more susceptible to burnout and have more difficulty solving customer problems.
Given these caveats, what might be the best way for agents and AI to collaborate moving forward?
New research from Stanford and MIT sheds some light. In what may be the first study of generative AI when used at scale in the workplace, researchers measured the productivity of more than 5,000 customer service agents at a Fortune 500 software firm over a year. Agents who used AI scripts generated from thousands of customer-agent interactions labeled as successful or unsuccessful boosted their productivity—measured as issues resolved per hour—by 14% on average. And improvement was even more striking for “novice and low-skilled workers” who were able to get their work done 35% faster.
Notably, the highest-skilled workers, whose tacit knowledge had been encoded in the AI at the outset, showed a small but statistically significant decrease in resolution rates and customer satisfaction. The paper’s authors speculated this might have been because agents could decide to accept or reject the AI-generated suggestions, creating an extra step for the highest-skilled workers who were doing their jobs well to begin with.
The study also revealed that assistance from AI reduced the number of requests for managerial help and improved retention. Said co-author Erik Brynjolfsson, director of the Stanford Digital Economy: “There was less churn once they used this tool because it seemed workers were happier and enjoyed the job more. We wondered if it would push them harder, but it seemed to be something workers liked. Customers were happier and I’m guessing as a call center operator, it’s more enjoyable to work with happy customers.”
Augmentation rather than replacement
It’s important to note that the AI system used in the Stanford and MIT study was designed to augment rather than replace human agents. It captured tacit “best practices” that were previously difficult for managers to articulate and was also able to provide more real-time recommendations than a busy manager.
In other words, the underlying assumption is that customer service is a distinctly “human” activity and that generative AI’s role is, first and foremost, to ease the load on real people whose job it is to help human customers. This includes protecting the cognitive and emotional resources—like judgment, flexibility, and empathy—required for moments of true interaction. Making this a guiding principle of adopting generative AI in call centers may quell concerns about the obsolescence of human agents and, paradoxically, help more of us, on both ends of the conversation, feel heard.