Key facts
- Vercel CEO Guillermo Rauch stated that AI agents are moving from prototyping to production, with coding agents and internal corporate agents being key use cases.
- Vercel has developed the Eve framework and Vercel Sandbox to manage agent instructions, skills, and data access securely.
- Rauch warned of risks associated with AI tools potentially training on sensitive company codebases without proper controls.
- Clients are increasingly adopting a plug-and-play approach to AI models, utilizing offerings from various providers like OpenAI, Anthropic, and Google's Gemini.
- Vercel aims to provide the infrastructure for AI agents, positioning itself as the 'AWS of this generation' by promoting open protocols and modularity.
Vercel CEO Guillermo Rauch believes the AI landscape is shifting from experimental prototyping to practical production applications, with a particular focus on coding agents and internal corporate agents.
Rauch highlighted the challenges of deploying AI agents in production, especially concerning secure data access and audit trails. To address these, Vercel has developed the Eve framework for defining agent instructions and skills, and Vercel Sandbox, a tool that isolates agents to control their data access and prevent sensitive information, such as proprietary codebases, from being used for training by third-party tools.
He explained that internal corporate agents can significantly boost productivity by providing employees, like sales representatives, with immediate access to data that was previously difficult to obtain. Rauch suggested that the rise of agents will compel many SaaS companies to open up their data, challenging traditional business models that rely on data enclosure.
The Vercel CEO also observed a change in how clients engage with AI labs, moving away from exclusive partnerships towards a more modular, plug-and-play approach. Clients are now selecting models from various providers, including OpenAI, Anthropic, and Google's Gemini, with Gemini and open-source models like Deepseek and GLM-5.2 gaining popularity due to their price-performance characteristics.
Rauch acknowledged that as AI labs expand their capabilities, they may directly compete with infrastructure platforms like Vercel. He emphasized Vercel's vision to be the 'AWS of this generation' for AI agents, advocating for open protocols and a software engineering-like approach where models and agents are distinct, modular components rather than tightly coupled entities.