BCG trains Jamie on top and worst sales behaviors

Business Insider reports that the Boston Consulting Group is training an AI agent named Jamie on both its top-performing salespeople and on examples of poor sales behaviour. Japjit Ghai, a managing director and partner at BCG X, said on the firm's podcast, "We trained the agent by studying the best sellers - their call transcripts, how they engage with customers - and teaching Jamie to do the same." Ghai added in an interview with Business Insider that BCG uses several sources to train Jamie, including internal expertise, a client's domain knowledge, and existing sales call recordings and transcripts, which he called a "repository of often underleveraged assets."

Companies building customer-facing agents often combine positive exemplars and negative examples to shape agent behaviour. Using recordings and call transcripts as training data aligns with imitation-learning approaches and supervised fine-tuning, where labelled good and bad interactions help define policy boundaries. From a data-practitioner standpoint, that requires careful labeling, versioning of dialogue datasets, and controls to avoid delegating harmful or noncompliant behaviours present in historical logs.

Industry reporting frames this as part of a broader trend where consultancies and enterprises convert frontline knowledge into reusable AI assets. Observers note that using internal sales conversations can accelerate domain adaptation for agents but raises privacy, consent, and compliance questions when customer data appear in training sets.

For observers and practitioners, monitor how teams operationalise negative-example curation, the metrics used to evaluate avoidance of "worst" behaviours, and any public statements from client organisations about consent and redaction practices. Also watch whether similar deployments publish technical evaluations or benchmark results that quantify behavioural improvements.