Vercel Cuts 10-Person SDR Team to 1 at $5,000 Annual Cost
Vercel reduced its lead qualification function from 10 people to one full-time US employee plus partial Europe and APAC coverage using an agent that costs $5,000 per year.
Vercel launched a lead qualification agent in August 2025 that reduced a 10-person SDR function to one full-time employee in the US plus 20 percent of one person covering Europe and APAC. The agent costs $5,000 per year in infrastructure and tokens and requires 20 percent of one engineer to maintain. According to SaaStr, the change delivered a reported 32x ROI by replacing 10 salaries with compute while running 24/7.
Lead Qualification Agent Results
The lead agent began as 20 percent of a single engineer’s time. After six weeks of human-in-the-loop refinement, it reached production scale. Vercel moved the original 10 people into higher-value roles rather than eliminating positions. SDR quotas increased 30 percent in the quarter after deployment across 30 workflows including event follow-up, product-qualified-account flows, and time-based campaigns.
GTM Engineering Team Formation
Vercel COO Jeanne DeWitt Grosser joined after running go-to-market at Google and Stripe for roughly a decade each. Six weeks after starting in June 2025 she created a go-to-market engineering team tasked with bringing agents to every GTM function. Ten months later the team had placed multiple agents into production, including a customer support agent handling 93 percent of total case load and a content agent completing 96 percent of major content updates last quarter.
Agent Build Process
Every internal agent follows the same three-person structure: a GTM engineer, a data scientist, and the single best subject-matter expert for the function. The team first documents the existing manual process in full, encodes it as deterministic workflows, then runs the agent in shadow mode with continuous human QA. For the lead agent a GTM engineer shadowed the top SDR across every tab and system for days before any AI component was added. According to SaaStr, the agent now matches 90th-percentile rep performance on every output.
Architecture Requirements
Vercel requires agents to operate through headless, composable interfaces rather than UIs. The Deal One meeting intelligence agent ingests Gong calls, generates notes and action items, posts Slack coaching, updates Salesforce fields, and tracks competitive mentions because both Gong and Salesforce expose accessible APIs and webhooks. The Playbook Platform triggers on usage signals or pricing-page visits, matches plays, and surfaces personalized outreach for one-click rep review. Tools lacking these surfaces were removed from the stack.
