Vercel CEO Guillermo Rauch says the AI sector is moving beyond experiments and into a more practical phase, where reliability, security, and workflow design matter as much as model power. At Vercel, that shift is already visible: the platform handles millions of deployments each day, and a large share now comes from coding agents.
Rauch describes two standout uses for agents. The first is software creation, where coding assistants are generating more output than ever. The second is the internal enterprise agent, built to help teams work faster by turning company data into immediate action. In both cases, the challenge is not intelligence alone, but safe access, traceability, and control.
To address that, Vercel developed Eve, a framework that lets teams define an agent's instructions and skills in natural language. The company also introduced Vercel Sandbox, a controlled environment that limits what data an agent can access or export while still allowing it to operate effectively. Rauch says this kind of structure is essential as businesses adopt AI tools across sensitive workflows.
He also argues that the market is becoming more modular. Instead of tying everything to one model provider, companies are increasingly combining models, gateways, data layers, and sandboxes as separate building blocks. That approach, he says, is driving interest in multiple AI systems, including Gemini and open models such as Deepseek and GLM-5.2, as teams optimize for production performance and cost.
For Rauch, the bigger shift is strategic: AI agents are pushing software companies toward openness, interoperability, and flexible infrastructure. As more products are built this way, the next phase of AI may look less like a single platform and more like a connected ecosystem of specialized tools. The future of AI could be shaped by how well these systems work together.