For ten years, Vercel's north star was simple: help a developer take an idea and turn it into a live URL as fast as possible. Then Guillermo Rauch saw a future where the entity on the other end of that URL wasn't always a person anymore.

Today, roughly 30% of deployments on Vercel are made by coding agents. Just a few months ago, that number was in the low single digits.

We brought a group of our portfolio founders together for a closed-door session on navigating the AI inflection point, and Guillermo was naturally one of the first people we called. Vercel is behind some of the most traffic-heavy deployments on the internet, and he's as close as anyone to the front lines of how software is actually changing.

The short version: the entity Vercel spent a decade building for has shape-shifted. Here's how Guillermo is thinking about what comes next.

The thing Guillermo kept coming back to throughout the conversation was ego death. He loved coding and built his career around it. He created frameworks that millions of developers use every day. And he still stood in front of his company to say that the future of coding is generation.

Where companies used to ship pixels (like a UI someone designs in Figma, implements once, and leaves largely static), they'll increasingly ship tokens. Inputs, outputs, background agents doing work continuously, whether or not a human is watching.

So Vercel fundamentally rebuilt its business around a consumption model.

Teams using Claude are already deploying 7.5 times more frequently than teams that aren't. "The agents might use more infrastructure, less infrastructure, more tokens, fewer tokens," he said. "That’s a new paradigm."

In Guillermo's exec meetings, there's one question he's stopped asking almost entirely: “What's the answer?” Instead, he asks: “What’s the agent that can get me this answer?”

This might sound like semantics. It isn't. It forces every leader to think in systems rather than outputs — to build something that scales instead of something that runs once. It's also how he knows whether his team is actually operating in the new paradigm or just talking about it.

At Vercel, it's already playing out across the business: the data team no longer answers data questions, they maintain the agent that answers questions from anyone in the company in Slack. Support engineers no longer resolve tickets, they curate the agent that resolves 90% of them (with a path toward 100%).

His bar for the company, from the IC layer through his executive bench, is 100% agent proficiency. The difference is what that means at each level — ICs ship with agents, leaders build the systems that make it possible at scale.

Vercel has always been known for highly curated, high-conviction products. That reputation is an asset — but it also became a constraint. AI moves fast enough that by the time you're certain, you're already late.

Guillermo's solution was to carve out a separate unit, Vercel Labs, with an explicit remit to be experimental, while keeping core infrastructure moving at the pace enterprises require. The two can't run at the same speed, and pretending otherwise slows both down.

"I wish I had arrived at this separation a little sooner," he told us.

Which, coming from Guillermo, lands differently.

“I'm a pretty impatient person," Guillermo said. "I just want to see things happen right now." It's why Vercel exists in the first place, he didn't want to set up AWS, he wanted a URL, instantly. The agent era is that same relentless instinct at a wildly different scale.

Guillermo isn't the only one rethinking who (or what) is on the other end of the internet. In a recent episode of Notable Perspectives, Chris Andrew, co-founder and CEO of Scrunch, makes the case that the same shift is coming for marketing.

"AI agents are working on behalf of a human and if I don't understand how to relate to that agent and market to the agent, I'm missing the human,” Chris told us.

In this episode, Chris and Notable investor Chelcie Taylor get into what it actually means to market to an agent instead of a person.

Listen Now: YouTube, Spotify, Apple