Vercel Cuts SDR Team with $5K‑a‑Year AI Agent

Vercel’s chief operating officer, Jeanne DeWitt Grosser, revealed at SaaStr AI that the company slashed its inbound‑lead SDR team from ten people to a single AI‑driven lead‑qualification agent that costs under $5,000 a year in infrastructure and token usage. The move lifted SDR quotas by 30% that quarter and now resolves roughly 93% of support cases with an internal AI agent.

What happened

According to SaaStr, Vercel built an internal lead‑qualification agent in roughly 20% of an engineer’s time and launched it with a human reviewer. Six weeks later the ten‑person SDR squad was reduced to one person in the US and a part‑timer covering Europe and APAC. The agent “performs like a 90th‑percentile rep on every lead,” Grosser said, and its cost is “under $5,000 a year in infrastructure and tokens.”

The company applied the same three‑person workflow to 30 SDR processes—including event follow‑up, product‑qualified‑lead flows, and time‑based campaigns—resulting in a 30% quota increase that quarter. Vercel also built a support agent called Vertex (dubbed “our internal Sierra”) that now handles about 93% of support tickets, replacing a vendor‑run solution that required 150 engineers.

All agents are built in‑house, side‑by‑side with top performers who document their workflows. The agents call APIs and webhooks (e.g., Gong, Salesforce) rather than clicking through UIs, so any tool lacking a developer surface would be dropped, Grosper noted. The meeting‑intelligence agent ingests sales calls, writes notes, suggests coaching, updates CRM fields, and surfaces competitive insights—all within Slack, with no separate UI.

A company‑wide data analyst agent, d0, translates plain‑English questions into SQL against a semantic layer that Vercel’s head of data science built. Queries that once took a week now return in under a minute, and every other agent leverages this layer to avoid hallucinations and generate richer copy.

Why it matters for small‑team growth hackers

For founders and GTM leaders running lean teams, Vercel’s playbook proves that a weekend prototype, maintained by a fraction of an engineer’s time, can outperform a full‑stack vendor solution. The math is stark: $5 K a year versus ten full‑salary SDRs, plus a 30% lift in quota. If your data warehouse and semantic layer are already in place—or can be built in a few weeks—you can replicate the same “agent‑first” stack without waiting for a multi‑year vendor contract.

The key takeaways are: (1) headless, API‑first tools are non‑negotiable; if a SaaS product can’t be called programmatically, Vercel would replace it. (2) High‑quality, enriched data is the foundation of reliable agents—Vercel’s agents only work because the semantic layer feeds them accurate, causal metrics. (3) Human oversight remains essential; agents run with a reviewer who steps in when the expert’s workflow is exhausted, ensuring quality while freeing reps for higher‑value work.

What to do this week

1. Identify one repetitive SDR workflow (e.g., event follow‑up) and map the exact steps a top performer takes; then prototype a simple script that calls your CRM and email API to automate it.
2. Set up a Slack‑based query bot that connects to your existing analytics or BI tool via an API, so reps can ask plain‑English questions and get instant answers.
3. Audit your current SaaS stack for API coverage; flag any tool without a public API and start researching a replaceable, headless alternative.

Sources

• Top 5 Learnings from Vercel’s COO: 10 SDRs Down to 1, a $5K-a-Year Lead Agent, and 93% of Support Cases Handled by AI — SaaStr