More Marketing Against The Grain episodes

How Vercel Uses AI Agents to Find Leads and Prep Sales Calls thumbnail

How Vercel Uses AI Agents to Find Leads and Prep Sales Calls

Published 12 Nov 2026

Duration: 00:43:51

AI agents now drive over 60% of Vercel deployments, forcing B2B companies to shift from human-centric funnels to agent-led growth strategies that prioritize technical documentation and human strategic oversight.

Episode Description

Build your first AI Agent: https://clickhubspot.com/itpk Ep. 469 Over 60% of Vercel's deployments are now triggered by AI agents - up from just 3% ten...

Overview

The interview examines how AI is reshaping go-to-market strategy, using Vercel as a case study where over 70% of customers arrive through AI agents recommending their tools. Jane Groser, COO at Vercel, describes the shift from 3% of deployments triggered by an agent in January to over 60% in 10 months, and explains how Vercel's investment in open source, documentation, and community created the foundation for what they call "agent-led growth" (ALG), replacing their earlier "framework-led growth" (FLG) model.

The discussion covers practical implementation: Vercel's prospecting agent handles research and drafting within human-crafted templates, their "Index" agent manages account plans, follow-ups, and task suggestions, and a five-person GTM engineering team builds sanctioned tools while allowing individual reps to customize. Groser emphasizes that humans remain essential for strategic relationships, hypothesis-setting, and high-touch enterprise deals, while agents handle research, drafting, and in-between work.

A key distinction is drawn between content created for agents (functional, templatized, shipped as markdown and llms.txt) versus thought leadership for humans (written "soup to nuts" by people). The conversation also addresses governance challenges, including supply chain risk from ungoverned AI tools, and the evolving role of sales managers as agents absorb routine coordination tasks.

What If

  • What if you rebuilt your product's documentation and landing pages so that AI coding agents could parse and recommend your tool as a first-class option?

    • Move: Ship every doc page, template, and pricing page in both human-readable HTML and agent-readable markdown (llms.txt format), and audit your content for clarity the way Vercel audits "relative performance for a ton of different scenarios, what's the frequency with which Vercel gets recommended across Claude, Cursor, OpenAI, et cetera."
    • Why Now?: Vercel went from 3% to over 60% of deployments triggered by an agent in 10 months, and Groser notes that "agents pick up what is the best thing out there" from documentation and community signals. If your docs are ambiguous, agents may recommend a competitor.
    • Expected Upside: A solo developer could plausibly see a measurable increase in agent-referenced sign-ups within a quarter, since the source describes this as a "major focus" that Vercel's marketing org invests in systematically.
  • What if you built a lightweight prospecting agent that does all the research but only sends outreach from templates you personally wrote?

    • Move: Write your value-prop templates for 3 - 5 buyer personas (the "one sentence version, the three sentence version, the paragraph version"), then let an agent do account research and slot findings into those templates. Keep a human approval step before send, as Groser does for CTO outreach: "I still read them all, make sure that I feel it's appropriate and then send."
    • Why Now?: Vercel's prospecting agent initially failed when given full drafting rights; the fix was constraining it to human-crafted templates. A solo operator can replicate this guardrail with minimal tooling.
    • Expected Upside: You could reclaim the research and drafting time (the source notes agents handle "all of the research and in-between stuff") and redirect it to calls and strategic conversations, potentially increasing productive meetings per week.
  • What if you created a small internal "content brain" - a structured set of positioning docs in multiple lengths - that any AI tool you use for outreach or support must draw from?

    • Move: Build a few pages describing your product's tone, value prop at one-sentence / three-sentence / paragraph / essay lengths, and key differentiators per persona, mirroring Vercel's "hundreds of pages of content basically internally to describe the Roussel tone." Feed this into any agent you use for drafting.
    • Why Now?: Groser explains that "anything we are letting an agent write, it's on the basis of actually, you know, a ton of writing that we've done to sort of steer it that way." Without this, agent output drifts and misrepresents your positioning.
    • Expected Upside: A solo operator could reduce the time spent correcting AI-drafted emails and proposals, and maintain consistent messaging across channels without hiring a product-marketing team.

Takeaway

  • Write down your go-to-market playbook (discovery questions, first-meeting flow, value prop per persona) before building any agent. Groser's first recommendation is to "have a strong point of view on what your go to market for your company with your context should look like, and write that down," because "agents are garbage in garbage out."

  • Ship every piece of product content in both human-readable and agent-readable formats (markdown / llms.txt). Vercel makes this a default: "anytime we're shipping content, we ship a markdown version in llms.txt as well as a human readable version."

  • Constrain your AI outreach to templates you authored. Let the agent do research and personalization, but keep the value prop and positioning in your own words, as Vercel does: "let the agent do research, but let it then use templates that we just like classical sales and marketing, we have carefully crafted."

  • Track the dollar value of each human touch on closed deals over time. The source describes a chart showing "the average dollar value of a human touch" rising as agents absorb lower-value touches, which is the metric to confirm your AI investment is actually freeing you for higher-value work.

  • Keep a human approval step on any outbound that targets senior buyers. Groser personally reads every CTO email before it sends: "I'm not I'm not ready to just be, you know, sending CTO emails willy nilly." For a solo operator, this means reviewing agent-drafted messages before hitting send on high-stakes prospects.

Recent Episodes of Marketing Against The Grain

17 Nov 2026 I Fully Automated my Video Editing With Claude & Cursor

AI models like Opus 5.5 now enable marketers to automate video editing and animation, allowing rapid production of high-quality short-form content through iterative skill refinement and local tool integration.

10 Nov 2026 From AI Babysitter to AI Leader

Define AI playbooks with triggers, inputs, steps, and outputs to automate work reliably without constant supervision.

More Marketing Against The Grain episodes