Vercel trained an AI agent on its best salesperson. Then it cut the 10-person team down to 1. (www.businessinsider.com)

Vercel announced it has built AI agents by shadowing its top-performing sales rep and automating much of the team’s inbound lead work, shrinking a 10-person sales development group to a single human overseer plus a bot. Engineers spent six weeks documenting the top performer’s workflows, then trained agents to review inbound messages, filter spam, query internal databases, research company details using OpenAI’s Deep Research tool, draft personalized responses, and route support inquiries. A human manager reviews the agent’s Slack outputs and provides feedback to refine tone and accuracy. Vercel has six agents in production, aims for hundreds in 6–12 months, and recently hired Databricks alum David Totten to lead the effort. This approach matters because it shows enterprises can codify tacit, repeatable human expertise into autonomous agents to accelerate routine work and redeploy people to higher-value roles. Vercel emphasizes agents are best suited to “replicable and deterministic” tasks, leaving ambiguous, creative, and complex responsibilities to humans; nine displaced reps were moved into outbound prospecting. The project illustrates a practical production pattern—human shadowing → workflow documentation → agent automation → human-in-the-loop feedback—highlighting efficiency gains, scalability, and governance questions around dependence on external LLM tooling, continuous retraining, and workforce impacts.