How Vercel Runs on AI Agents: 96% of Marketing, 93% of Support, and an SDR Team Reabsorbed. A Deep Dive With CPO Tom Occhino
Summary
Vercel's Chief Product Officer, Tom Occhino, shared insights at SaaStr AI Deploy on the company's extensive internal deployment of AI agents, achieving significant operational efficiencies. Vercel utilizes hundreds of agents, with a content agent generating 96% of marketing content drafts and a customer support agent handling 93% of inquiries without human intervention. A lead qualifying agent also enabled the redeployment of an entire SDR team. Occhino emphasizes that the future of software is shifting from UI-first applications to autonomous, headless agents, and that "agents as a service" are ineffective. Instead, companies should invest in robust foundational infrastructure and tools for building custom agents. Vercel's "DealOne" go-to-market agent, which processes sales calls for intelligence and coaching, exemplifies this approach, built on their proprietary "agent stack" including AI SDK, AI Gateway, and Workflow SDK.
Key takeaway
For AI/ML Directors or entrepreneurs planning agent deployments, recognize that building custom agents on robust infrastructure, not buying "agents as a service," is key. Focus your team's energy on documenting repetitive, low-value tasks meticulously, as this is the true bottleneck. Then, leverage coding agents to build and iterate these foundational automation tools, freeing your human talent for high-impact, compounding work. Prioritize reliable infrastructure like durable workflows and fallback routing for production readiness.
Key insights
Vercel's experience shows autonomous AI agents, built on robust infrastructure, drive significant operational efficiency by automating repetitive tasks.
Principles
- Most agent-related development is undifferentiated "heat loss."
- Agents shift software from UI-first to autonomous, headless operations.
- Reliability infrastructure is mandatory for production-grade agents.
Method
Identify low-value, repetitive work. Document the process in detail. Use a coding agent (e.g., Claude Code, v0) to build and iterate the agent to production.
In practice
- Automate support triage, data pulls, or lead qualification.
- Use AI SDK for model integration and seamless switching.
- Deploy agents in secure, isolated environments like Vercel Sandbox.
Topics
- AI Agents
- Vercel Platform
- Agent Infrastructure
- Workflow Automation
- Go-to-Market AI
- Autonomous Software
Best for: Director of AI/ML, AI Engineer, Entrepreneur
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by SaaStrAI.