There’s a lot of public concern that an AI will someday go off the rails and annihilate humanity.<\/p>\n
\n Loading audio narration…\n <\/p>\n
So it may be a little surprising that Boston Consulting Group<\/a> is training its customer-facing AI agent, Jamie, on what not to do.<\/p>\n Japjit Ghai, a managing director and partner at BCG X, the firm’s tech build-and-design division, said on a recent episode of the comany’s podcast that Jamie is learning from both its strongest-performing sellers \u2014 and their worst decisions.<\/p>\n “We trained the agent by studying the best sellers \u2014 their call transcripts, how they engage with customers \u2014 and teaching Jamie to do the same,” he said.<\/p>\n “We also trained Jamie not to replicate the worst seller experiences.”<\/p>\n BCG trains Jamie using several sources, including its own internal expertise, a client’s knowledge of its business, and a company’s existing sales calls and transcripts, Ghai later told Business Insider in an interview.<\/p>\n Those recordings and transcripts are a “repository of often underleveraged assets,” Ghai said, that BCG can use to identify “what good looks like.” The goal is not to mimic one individual seller, he said, but to use historical patterns to understand which behaviors and ways of engaging with customers “tend to resonate better.”<\/p>\n Jamie is also trained to identify behaviors that “did not resonate.” BCG can then use those lessons for targeted coaching and development. After a call, he said, a seller could get a personalized scorecard<\/a> from Jamie showing what they did well and where they fell short, based on how they navigated the conversation.<\/p>\n Ghai said the system is designed to improve over time. Each conversation creates more data that can be used to train future conversations, he said, making Jamie an “always on muscle” that learns from the best individual behaviors rather than simply copying the best individual sellers.<\/p>\n BCG isn’t the only company to use its top talent for training purposes.<\/p>\n At Vercel, a cloud-based platform for developers to build and deploy websites and applications, agents have also been trained on their best employees. Last year, the company modeled an AI agent<\/a> after its best sales development representative and ultimately downsized its 10-person team to one person overseeing the agent.<\/p>\n “Modeling after the top-performing employees has always been a standard business practice. The difference now is that technology lets us accelerate it,” David Totten, Vercel’s vice president of global field engineering, told Business Insider in October.<\/p>\n