Vercel CEO Guillermo Rauch sees a clear dividing line forming in the AI industry: one side wants to bundle the model and the agent into a single, closed product, while the other—his side—wants to keep them separate. In a conversation following the company’s ShipNYC conference last week, Rauch laid out his vision for a modular AI future, where developers can swap models like they swap databases, and where infrastructure platforms like Vercel become the central nervous system for AI-powered applications.
Vercel, best known for its cloud platform that lets developers deploy code without managing servers, has quietly become a critical piece of the AI infrastructure puzzle. The company now processes 6 million deployments per day, half of which are triggered by coding agents, and more than 1 trillion tokens flow through its AI gateway daily. Those numbers underscore a shift from last year’s experimental phase to a focus on production-grade AI.
Also read: UK regulator warns of ‘arms race’ to keep up with AI use in financial services
From prototyping to production: The two killer apps of agents
Rauch described 2024 as a year of rapid experimentation within Vercel itself, with hundreds of agents developed and deployed internally. The real lessons came when those agents hit production. “The biggest lesson for me was the home-run use cases, the two killer apps of agents,” he said. The first is the coding agent, which drives massive token utilization but also generates an enormous volume of software that needs a home. The second, he argues, is the internal corporate agent—tools that help run the company by securely accessing data, auditing actions, and maintaining a trail of tool calls and access controls.
To address the security and governance challenges of these internal agents, Vercel built a framework called Eve, where agent instructions and skills are defined in natural language. The company also introduced Vercel Sandbox, which places agents in a restricted environment. “It can have the freedom still to do to express its intelligence, but then you can apply policy on what data it can access and what data can leave the sandbox,” Rauch explained. He cited a real risk: a developer tool like Devin or Cursor, if misconfigured, could train on an entire company’s proprietary codebase, a concern he says was raised directly by the president of Airbus.
The rise of the internal corporate agent
Rauch offered a concrete example of how these internal agents change day-to-day operations. A sales representative at Vercel, responsible for growing existing accounts, previously faced a data bottleneck. “She couldn’t ask that question in the past. She needed to wait until a Q1 project for a new sales dashboard completed,” he said. Now, with Eve, that same sales rep can ask in real time for the five accounts that added the most seats in the last two weeks, and get an answer immediately. “On the R&D side, we’re the fastest-moving company in the world. But on the sales engine, the Salesforce engineering [side], I was so incompetent. I had never opened Salesforce in my life when I started,” Rauch admitted. “Now I feel like I can actually have impact across the entire company.”
Open ecosystems vs. walled gardens
Rauch sees agents as a force that will force companies to open up their data. “So many of these SaaS giants build their entire kingdoms on trapping your data, and that’s incompatible with agents,” he said. He noted that clients are moving away from picking a single AI lab partner and toward a plug-and-play approach. “You can use OpenAI, you can use Anthropic, or you can use Gemini. We’re seeing a lot of growth of Gemini, even though it’s not on the news as much, because people are optimizing for production now.” He also highlighted the rise of open models like DeepSeek and GLM-5.2, which are gaining traction due to their price-to-performance characteristics.
This puts Vercel in direct competition with the major AI labs. OpenAI recently released tools that allow users to publish content directly to the web without leaving its ecosystem. “It’s a natural next step for them to host little websites,” Rauch said. “And it’s a great opening for us, because now people will think of ChatGPT as a tool for making websites. And then if they keep asking the model questions about web hosting, the model recommends us.” He framed the broader conflict as a choice: “Do you get all your intelligence from one place? Or do you get a module or a library or a building block from one provider, and then you build on top of it. That’s more like software engineering has always been, and that’s really what we’re bringing to market. We’re going to be the AWS of this generation, so obviously we’re fighting for a world of open protocols.”
Frequently Asked Questions
What is Vercel’s role in the AI industry?
Vercel provides cloud infrastructure that allows developers to deploy AI agents and applications without managing servers. It processes 6 million deployments per day and over 1 trillion tokens through its AI gateway.
How does Vercel’s ‘Sandbox’ feature protect company data from AI agents?
Vercel Sandbox places an AI agent in a restricted environment where administrators can apply policies on what data the agent can access and what data can leave the sandbox, preventing unauthorized training or data leaks.
Why does Guillermo Rauch believe models and agents should be separate?
Rauch argues that keeping models and agents as separate, interchangeable components allows developers to choose the best model for each task, promotes competition, and prevents vendor lock-in, which he believes is healthier for the software ecosystem.
