muzzdang

Second Brain notes from YouTube & builds

Second Brain notes

Marketing & Lead Gen: How to Get AI Agents to Recommend Your Business (Vercel COO)

Watch on YouTube: https://www.youtube.com/watch?v=Z0IjcAYzYBQ

Channel: Marketing Against the Grain

Vercel’s Shift to Agent-Led Growth: What the Numbers Show

Vercel COO Jeanne Grosser recently shared how the company moved from a framework-led growth model to an agent-led growth (ALG) model. The headline figures are striking: over 70% of new customers now arrive via AI coding agents such as Cursor, Claude Code, and Lovable, and agent-triggered deployments jumped from 3% to 60% in ten months (January–October 2024). ChatGPT-driven signups rose from 1% to 10% in the same period. These are self-reported metrics from Vercel’s leadership, reflecting a developer-tool category where AI coding agents have unusually high influence.

What the Conversation Covers

The discussion walks through three interconnected layers: how Vercel restructured its content and product surfaces for agent consumption, how it built a governed GTM engineering function to create sanctioned sales agents, and how it rewired sales roles and quotas around the time AI returns to reps. Grosser frames the shift as “capability-driven product development” — every feature must be accessible via CLI, MCP, and dashboard before launch because agents are now first-class buyers.

Step-by-Step Walkthrough of Vercel’s Approach

1. Content Engineering for Agent Ingestion

Vercel publishes every piece of technical content — docs, templates, guides — in both human-readable HTML and llms.txt (markdown) formats. The marketing team treats agents as a primary distribution channel. A curated “sales agent brain” of value-prop copy feeds agent-written functional content, while thought-leadership pieces remain human-written. Grosser’s rule: “never outsource your taste to AI.”

2. Product Exposure via CLI, MCP, and Dashboard

No feature ships until it is available to an agent through the command line, the Model Context Protocol (MCP), and the web dashboard. This ensures that when a developer asks an agent “where should I deploy this Next.js app?”, Vercel appears as the default, well-documented choice.

3. GTM Engineering Squad (5 People)

A dedicated team builds sanctioned internal agents: a headless CRM front end (100% daily active usage), a RevOps agent (“Revoa”), a MOPS agent (“Mop rater”), and the prospecting agent “Index.” Centralizing API keys, auth, and permissions prevents “AI psychosis” — reps building duplicate, insecure tools on their own.

4. Prospecting Agent “Index”

Index handles research, builds 90% of account plans, drafts call summaries, follow-up emails, and task nudges. Reps approve and send. The goal is to move BDRs from email volume to high-value human conversations. Sales managers shift from call review to strategic deal involvement and GTM R&D; Vercel plans to increase the rep-to-manager ratio because AI now handles forecasting and call analysis.

5. Quota Adjustments and Pilot-Driven Rollout

Grosser raises sales targets every six months (done for 18+ months). If AI returns ~40% of touches, quota must rise or headcount adjusts. The rollout follows a pull strategy: one team co-builds the agent with GTM engineering, proves lift, then expands.

Key Numbers and Tools Mentioned

  • 70% of customers via AI agent recommendations
  • Agent-triggered deployments: 3% → 60% in 10 months
  • ChatGPT signups: 1% → 10%
  • 3-person MOPS team supporting the entire organization
  • 5-person GTM engineering function
  • 100% daily active usage of headless CRM front end
  • 2–3 CTO meetings per day for the COO (10–15 per week)
  • 90% of account plans built by agent
  • Tools: Vercel, Next.js, Cursor, Codex, Claude Code, Lovable, Salesforce, Snowflake, Gong, MCP, llms.txt, CLI

Caveats and Hype Check

The 70% and 3%→60% figures are self-reported by Vercel’s COO with no independent audit; they reflect a developer-tool category where AI coding agents are unusually dominant. The claim that “nearly every go-to-market is wrong” is rhetorical framing. No affiliate links or revenue promises to viewers. HubSpot sponsorship appears only in the closing ad read.

Enterprise buying dynamics haven’t fundamentally changed — buyers still minimize risk — but they lack established requirements for AI stacks, making sales more consultative than RFP-driven.

Concrete Next Actions for Your Team

  • Audit and document your GTM playbook: discovery questions, meeting agendas, value props per persona, follow-up sequences.
  • Publish all technical content in dual format: HTML and llms.txt (markdown) so coding agents can ingest it.
  • Expose every product capability via CLI and MCP before dashboard: treat agents as first-class users.
  • Create a small GTM engineering squad: build sanctioned agents for prospecting, RevOps, MOPS with governed data access.
  • Implement hypothesis-driven account planning: reps write the strategy; agents execute research and drafting.
  • Track “productive meetings per week” as the north-star sales metric: automate prep, follow-up, and forecasting.
  • Raise quotas proportionally to time saved by AI: redeploy or exit reps who won’t adopt the new motion.
  • Govern AI tool sprawl: centralize API keys, enforce auth/permissions, deduplicate agent builds across teams.
  • Separate content streams: human-written for thought leadership; agent-written (from curated “content brain”) for SEO/AEO.
  • Run a pilot: pick one sales team, co-build an agent with GTM engineering, measure lift, then scale.

Final Thought

Vercel’s experience illustrates a broader pattern: when the buyer is an agent, the product and the go-to-market motion must be agent-native from day one. The prerequisite is not buying more AI tools — it is codifying your best practices so that agents have high-quality inputs. As Grosser puts it, “agents are garbage in, garbage out.” Start by writing down exactly how your best reps work today; only then can you automate the 80% of low-value work and free humans for the 20% that drives revenue.