BCG says it is training its customer-facing AI agent, Jamie, using both exemplary and poor sales behaviors to improve coaching and performance.
An AI agent learning how to sell better is drawing attention for a surprisingly practical reason: BCG says it is also teaching the system what salespeople should not do.
Boston Consulting Group’s customer-facing AI agent. Jamie. is being trained on customer service behaviors drawn from its clients’ sales teams. according to BCG’s tech division.. In remarks shared through the firm’s podcast. Japjit Ghai. a managing director and partner at BCG X. said Jamie is learning from both “best” and “worst” examples of how clients’ salespeople engage with customers.
Ghai said the training approach involves studying top-performing sellers by reviewing call transcripts and how they interact with customers. then teaching Jamie to mirror effective patterns rather than simply copying a single individual’s style.. At the same time. he said the agent is explicitly trained not to reproduce the poor experiences associated with the worst seller behaviors.
BCG trains Jamie using data its clients share so the system can analyze customer service calls. and it also relies on BCG’s own internal research.. The materials described include a client’s knowledge of its business as well as existing sales calls and transcripts already in use by the company that provides input into the training process.. In Ghai’s framing. these recordings and transcripts can function as a “repository” of underused assets that help define what “good” looks like in real interactions.
The company’s goal is not to replicate one salesperson. but to identify historical patterns in customer conversations—especially those behaviors and ways of engaging that tend to resonate better.. At the same time. Jamie is also designed to spot behaviors that “did not resonate. ” creating a dual pathway of reinforcement and correction based on what appears to work and what appears to fail in prior calls.
After each conversation, Ghai said Jamie can deliver feedback in a structured way.. In his description. a seller could receive a personalized scorecard highlighting what they did well and where they missed the mark. based on how they handled the interaction.. That feedback loop is positioned as a performance-coaching tool rather than a replacement for sales roles.
BCG also said the system is engineered to improve over time.. Each conversation adds more data. which can then be used to train future interactions. making Jamie an “always on muscle” that learns from strong individual behaviors without merely copying the best performer.. The implication is that the agent’s effectiveness can evolve as more customer-facing sessions are processed.
BCG is not alone in treating internal talent performance as a training input for AI.. Vercel, a cloud-based platform where developers build and deploy applications, has also trained agents by modeling them on top employees.. In earlier reporting. Vercel’s vice president of global field engineering. David Totten. said the company modeled an AI agent after its best sales development representative and later reduced its team from 10 people to one person overseeing the agent.
Totten framed that shift as an acceleration of something companies have long done: using top performers as benchmarks. The difference, he said, is that technology allows firms to replicate that learning faster than traditional coaching or process changes.
The business relevance of these moves lies in how customer interactions are increasingly treated as data.. If call transcripts and customer conversations are used to formalize effective behaviors. coaching can become more immediate and more consistent across a sales team. especially when the feedback is tied to specific parts of the conversation rather than broad guidance.
There is also a risk-management angle to the “worst behavior” element. even if the company does not describe it in safety terms.. By explicitly training against poor experiences. an agent can be shaped to avoid repeating patterns that customers tend to react to negatively—an approach that. in practice. turns lessons from past mistakes into part of day-to-day decision support.
Meanwhile. the use of ongoing learning suggests companies will increasingly measure sales quality through conversational analytics. not only through outcomes like revenue or conversion rates.. Over time. that could shift how organizations allocate training resources and how managers evaluate performance—more around communication behaviors that correlate with customer response.
A correction was also issued noting that BCG trained Jamie on data provided by its clients about customer service interactions.
BCG X Jamie AI agent customer service training sales coaching AI analytics calls enterprise AI sales development