How Vercel Cut Its Sales Team From 10 to 1

The one thing you need to know in AI today | AI Ready CMO

Oct 29, 2025

Yes, this is a story about cutting headcount—Vercel (a developer-focused hosting platform) just reduced its inbound sales team from 10 people to 1.

But instead of philosophizing about AI displacement, let’s get practical.

Buried in this news is a remarkably clear implementation roadmap that any B2B SaaS company could follow. And if you’re a senior marketer thinking about where AI agents fit in your operation, this is the most concrete case study we’ve seen for how to actually make the transition work.

The Implementation Playbook

Here’s what Vercel’s team actually did:

Identified their top performer: Not the most experienced rep, not the most senior—the one with the best measurable results
Shadowed for 6 weeks: Three engineers followed this star SDR, documenting every single action, decision point, and workflow step
Built the agent to replicate that exact process: The agent now reviews inbound messages, filters spam, qualifies leads using internal databases and external research tools (they use OpenAI’s Deep Research), drafts personalized responses, and routes support inquiries
Human-in-the-loop review via Slack: One person reviews the agent’s work and provides ongoing feedback to tune tone, improve decision boundaries, and maintain brand voice
Redeployed, didn’t fire: The other 9 team members moved to outbound prospecting—more complex, higher-value sales work

Why This Worked (And Why Most Attempts Won’t)

The success factors here are simple.

First, they picked work that’s “replicable and deterministic”—Vercel’s COO specifically used those terms. In other words, tasks where you can document a clear process and expect consistent outputs. Inbound lead qualification fits that perfectly. Complex enterprise negotiations with multiple stakeholders? Not so much.

Second, they trained the AI on excellence, not mediocrity. This matters more than people realize.

Most companies would compile “average” workflows from their entire team. Vercel deliberately modeled only their best performer—the same way you’d pair your strongest intern with your star employee, not your most available one.

Third, they kept human judgment in the loop. This isn’t full automation, but augmented execution with human oversight, maintaining quality and handling edge cases.

The Practical Reality Check

If you’re running SDR operations, customer support, or early-stage qualification in marketing, this model is now viable. The technology exists. The implementation path is clear. But you need high-quality training data (transcripts, emails, documented playbooks), you need to identify which workflows are truly deterministic, and you need someone who can provide continuous feedback to keep the agent aligned with your standards.

What won’t work: trying to automate complex objection handling, multi-stakeholder negotiations, or anything requiring significant judgment calls under ambiguous conditions. The agent handles the high-frequency, standardized work. Humans still own the complexity, creativity, and relationship-building at scale.

We know that entry-level sales roles as training grounds are shrinking. The new entry point might be “AI sales manager”—someone who can tune, improve, and work alongside these systems. That’s a harder first job to land, but potentially a more valuable skillset long-term. Whether that trade-off is worth it for your organization depends on how you value developing junior talent versus operational efficiency.

Vercel made their choice. What’s yours?

Found this useful? Share it with a colleague who needs to read this.

Gamma

Create polished presentations in minutes without design skills—Gamma AI generates slides, formats content, and suggests layouts automatically, turning outlines into professional decks instantly.

→ Try it here