What Happened
Vercel, the $9.3 billion developer infrastructure company, reduced its SDR team from 10 people to one after deploying a custom-built AI agent for inbound lead qualification. The company ran the agent in shadow mode for six weeks while its top-performing SDR reviewed every output. Once the human could no longer improve the agent's performance, Vercel made the switch.
COO Jeanne DeWitt Grosser reported the agent now performs at a 90th percentile level, 100% of the time. It costs under $5,000 annually in infrastructure and token spend. The remaining SDR's role shifted: they now train and refine the agent rather than qualify leads directly.
The Quota Reality
Vercel increased SDR quotas by 30% in the quarter following the automation rollout. That detail matters because it signals the company expects higher output per rep when automation handles volume work. One human SDR plus the agent is now expected to deliver more pipeline than 10 humans did previously.
The economics are stark: 9 SDRs at $80k base each is $720k in annual salary before OTE, benefits, or tooling. The agent replacement costs $5k.
The Build Approach
Vercel stood up a go-to-market engineering team six weeks after Grosser joined. The team's mandate: deploy AI agents across every GTM function. Results so far include a support agent handling 93% of case volume (mostly technical queries), a content agent completing 96% of major marketing updates last quarter, and the SDR agent.
The company built these tools in-house rather than buying platforms like Artisan or other AI SDR solutions. That approach requires engineering resources most sales orgs do not have, but it gave Vercel tight control over training data and output quality.
What This Means for SDRs
This is not a pilot. Vercel is a high-growth company with enterprise customers including OpenAI, PayPal, and Walmart. It recently closed a $300 million Series F and is estimated to be running at $340 million ARR. When a company at that scale cuts 90% of its SDR headcount and raises quotas for the remainder, the signal is clear: automation is moving from experiment to operating model.
For SDRs, the job is changing. Vercel's remaining SDR is not doing qualification work anymore. They are training an AI system. That is a different skill set: pattern recognition, prompt refinement, quality control. Worth asking in your next interview: is this an SDR role or an SDR trainer role? The comp should reflect that distinction.
For sales leaders, the math is simple but the execution is hard. Vercel had the engineering talent to build this in-house. Most companies will need to buy platforms, which means evaluating vendors on output quality, integration depth, and total cost. A $5k agent sounds cheap until you factor in the sales tech stack, data hygiene work, and ongoing training time.
The other question: what happens to the pipeline when you lose institutional knowledge from 9 experienced reps? Vercel ran shadow mode for six weeks specifically to transfer that knowledge into the agent. That is the right approach, but it assumes your top SDR has time to train an AI system while still hitting quota. Plan for that transition cost.
ANZ Context
Vercel does not appear to operate a dedicated ANZ sales office or disclose local headcount. The company's reach in Australia and New Zealand comes through global enterprise accounts and developer adoption rather than a regional sales team. That means local SDRs are more likely to see this impact through their employers adopting similar automation strategies than through direct Vercel hiring or restructuring in ANZ.
The broader question for ANZ sales orgs: if a US company at Vercel's scale can cut SDR headcount by 90%, what does that mean for local teams? ANZ enterprise software companies typically lag US counterparts by 12 to 18 months on GTM trends. This one is moving faster because the tools are cloud-based and the ROI is immediate.