Vercel CEO Guillermo Rauch on the Future of AI Agents
Vercel, known for its cloud infrastructure that lets developers deploy agents without managing servers, has become a key player in AI software. The company now processes 6 million deployments per day, half triggered by coding agents, and over 1 trillion tokens flow through the platform daily. Following the ShipNYC conference in late June 2026, TechCrunch sat down with CEO Guillermo Rauch to discuss the evolving AI landscape and how Vercel is navigating competition with major AI labs.
From Prototyping to Production: Lessons Learned
Rauch noted that 2025 was about prototyping, with everyone building agents. Vercel saw hundreds of agents developed internally, which revealed production challenges. The two killer app use cases emerged: coding agents (driving token usage) and internal corporate agents that help run the company. The main hurdles for internal agents are secure data access, auditing, and tracking tool calls.
Introducing Eve and Vercel Sandbox
To address these challenges, Vercel developed the Eve framework, which lets users define agent instructions and skills in natural language. The Vercel Sandbox acts as a secure cage for agents, applying policies on data access and egress. This prevents risks like unintentional training on proprietary code, a concern Rauch shared after speaking with the president of Airbus about protecting decades of aerospace C++ code.
Internal Agents in Practice
Rauch described a sales rep at Vercel who uses Eve to ask: “Give me the five accounts that have added the most seats in the last two weeks.” Previously, such a query required waiting for a Q1 dashboard project. Now, with Eve, the same technology that powers customer-facing agents also boosts internal productivity. Rauch believes agents will force companies to open up, challenging SaaS giants that trap data.
Client Relationships with AI Labs
Companies are moving away from single-lab commitments. They now mix and match models: OpenAI, Anthropic, Gemini, Deepseek, and GLM-5.2. Rauch noted strong growth for Gemini, driven by price/performance in production. Vercel also faces direct competition from labs like OpenAI, which recently released tools to publish web content without leaving its ecosystem. Rauch sees this as an opportunity, as users may turn to ChatGPT for websites and then seek hosting advice—leading to Vercel recommendations.
The Decoupling of Models and Agents
Rauch emphasized a fundamental split: “Are the model and the agent going to be coupled?” He advocates for modular building blocks, comparing Vercel to the AWS of this generation. The company is fighting for open protocols, allowing developers to choose intelligence from different providers and build on top, as software engineering has always done.
- 6 million daily deployments
- Half triggered by coding agents
- 1 trillion tokens daily
- Eve framework for natural language instructions
- Vercel Sandbox for secure agent execution