Vercel Cloned Its Best Salesperson Into an AI Agent and Cut a 10-Person Team to One

Vercel spent 6 weeks studying its top sales rep, built an AI agent mimicking that workflow, and collapsed a 10-person inbound team to 1 human supervisor. The displaced workers were reassigned, not fired — for now.

Vercel, the $9.3 billion developer-platform company founded by Guillermo Rauch, has completed what may become the template for white-collar AI displacement: shadow your best employee for six weeks, encode every step into an autonomous agent, then shrink the team from ten to one.

The mechanism is specific. In June 2025, Vercel's COO Jeanne DeWitt Grosser hired three engineers and formed an internal AI initiative within the sales division. The engineers spent six weeks observing the team's top-performing sales development representative, documenting every step of the inbound lead-handling workflow — message triage, spam filtering, internal database queries, company research via OpenAI's Deep Research, lead qualification, personalized response drafting, and routing support inquiries. They then built an agent that replicates this process end-to-end. The "lead agent" now handles the bulk of inbound sales development work that previously required ten people. A single human manager reviews the agent's output on Slack and provides feedback, allowing the system to learn Vercel's house style and improve over time. The remaining nine workers were moved to outbound sales roles — described by Grosser as "higher-value, more complex" work. Vercel currently runs six AI agents and plans to deploy hundreds within six to twelve months, all modeled on internal top performers. The company insists this is not a headcount reduction story. Total employee count has grown over the past year, and Grosser frames the shift as liberating humans for creative and intellectually challenging tasks. David Totten, the newly appointed VP of Global Field Engineering (previously at Databricks and Microsoft), echoes the line: the technology accelerates an old business practice of modeling behavior on top performers. But the structural logic is clear and worth naming. The method Vercel describes — identify deterministic, replicable workflows and hand them to agents — is a systematic framework for eliminating entry-level knowledge work. Grosser herself identifies the defining trait of agent-suitable work: "replicable and deterministic, meaning the same input always produces the same output." That description covers a vast territory of junior white-collar employment. The reassignment narrative deserves scrutiny. Outbound sales is indeed more complex than inbound triage, but the pipeline of workers who once entered sales through inbound roles is now blocked. If the entry ramp disappears, the talent pipeline for the "creative, high-value" roles dries up within a few years. This is the classic automation paradox: you need the junior roles to produce the senior workers who supervise the agents. Vercel's approach is notable for its specificity — not a vague "AI transformation" but a documented, replicable playbook. That makes it both more credible and more consequential. Every company with a sales development team is now looking at the same math: three engineers, six weeks, one agent, nine fewer seats.