What Vercel's AI agent playbook means for your GTM

What Vercel's AI agent playbook means for your GTM

Vercel's CPO Tom Occhino just went public on how the company runs: 96% of marketing handled by AI agents, 93% of support, and an entire SDR function absorbed into automated workflows. Read the full breakdown in How Vercel Runs on AI Agents: 96% of Marketing, 93% of Support, and an SDR Team Reabsorbed. What jumped out to me was not the percentages. It was the architecture underneath them. Occhino didn't automate tasks. He replaced whole functions with coordinated agent systems. That is a different move entirely.

This is not about Vercel's scale

Before you file this under "enterprise stuff, not relevant to me," consider what Occhino actually built before Vercel. He co-created React at Facebook. He knows how to design systems that work at both seed and scale. The patterns he's describing at Vercel are not exclusive to companies with hundreds of engineers. They are about deciding which cognitive work belongs to humans and which belongs to machines.

For post-PMF B2B SaaS founders, that question is more urgent than it looks. You've found product-market fit. You're spending real money on paid media. You probably have one or two people trying to run your entire go-to-market. The GTM debt accumulates faster than the revenue does. Competitor positioning shifts. Channel performance drifts. You are reading week-old data and calling it attribution. That gap is exactly where coordinated agents pay off, and it is a gap that starts well below Vercel's budget.

The SDR absorption is the most important part

Vercel didn't just buy a few AI tools for their sales team. They restructured where the top-of-funnel work happens. Their SDR function no longer exists as a human-staffed department. It exists as an agent layer that handles prospecting, qualification, and initial outreach at continuous speed.

For founders running $50K to $500K per month in paid media, this matters because SDR work and paid media work overlap heavily. Both are doing the same job: finding qualified buyers before they raise their hand. If your SDR function is burning $30K a month to do work an agent layer can do at a fraction of the cost and ten times the volume, that is a CAC problem. It shows up in your blended numbers whether you see it or not.

Your attribution model is probably wrong before the agent question matters

This is the part nobody says out loud. Companies automate GTM motions that are already broken. They speed up bad targeting. They scale misattributed spend. Vercel's Occhino is explicit about the foundation: you need clean signal before you automate anything.

In paid media, that means knowing which channels actually generate pipeline, not which ones show the most last-click conversions. At $300K/month, I watch blended CPA move 20-35% based on attribution methodology alone. The channel mix that looks right on a last-touch model often looks completely different on a first-touch or data-driven model. Automate the wrong signal and your agent runs hard in the wrong direction.

This is why a functioning gtm strategy review comes before agent deployment. Not the other way around.

96% automation requires knowing the 4% that stays human

The headline number is 96% of marketing automated. The underreported number is that the remaining 4% is doing the highest-leverage work. Positioning decisions. Messaging bets. Creative strategy calls. The judgment that requires context a model doesn't carry.

Founders who hear "96% automated" and think "I should automate everything" miss the point. The 4% is where the compounding happens. That is where you decide which competitor angle to hit this quarter, which channel deserves a budget increase, which segment is being underpriced. Those calls require someone who actually knows your business. Agents execute. Humans decide what to execute on.

Agent coordination beats point solutions

What Occhino built is not a stack of separate AI tools running in parallel. It is coordinated agents with defined roles, handoffs, and feedback loops. That structure matters more than any individual agent's quality.

The same principle applies to your GTM intelligence. A competitor monitoring tool and a separate SEO tool and a disconnected attribution dashboard do not compound. They produce reports you cross-reference manually on Friday afternoons. Coordinated agents that share context, feed positioning signals into channel strategy, and score trends against your current spend produce a different output entirely. This is the core design principle behind GTMVP's GTM strategy framework: eight specialized agents that operate as a system, not a stack.

What GTMVP is built for

GTMVP is eight agents running continuously: competitor mapping, positioning refinement, angle generation, channel scoring, trend surfacing, and three more layers underneath. The design is deliberately modular but the output is unified. Each agent feeds the others.

The reason I built it this way is exactly what Occhino describes. Point solutions create coordination overhead. You end up managing your tools instead of running your GTM. GTMVP removes that overhead by handling the intelligence layer so you are making decisions on current signal, not a snapshot from last month. For founders running real paid media spend, that difference shows up in CPA within 60 to 90 days.

What to do this week

• Audit your attribution model before you add any agent tooling. Identify whether your current reporting is last-touch, first-touch, or something else, and what that model is hiding.
• Map your SDR-to-paid overlap. Calculate what percentage of your pipeline could have been sourced by a well-targeted paid channel instead. That number tells you where agent leverage is highest.
• Score your top three channels by true contribution to closed-won deals, not by impressions or clicks. If you cannot do this today, that is the blocker to fix first.
• Identify the 4%. List the GTM decisions that require your judgment specifically and cannot be delegated. Everything else is a candidate for the agent layer.
• Run a competitor positioning sweep this week. Vercel's agents are doing this continuously. Most of your competitors are not. That gap is your window.

If you want to see what a functioning GTM intelligence layer looks like against your actual business, run a GTMVP audit at /audit. You can also review a sample report before you commit to the full run. The gap between where your attribution is now and where it needs to be is usually smaller than it looks. Start there.