What we learned building agents at Vercel

Vercel shares a practical methodology for identifying and building successful AI agent projects. The key is targeting tasks with low cognitive load and high repetition—work that's too dynamic for traditional automation but predictable enough for current AI models. Two case studies demonstrate this approach: a lead processing agent that reduced a 10-person team to 1, and an anti-abuse agent that cut ticket closing time by 59%. The methodology involves asking teams about their most repetitive, mind-numbing tasks, then building agents with human-in-the-loop validation. Vercel provides open-source templates for lead processing, data analysis, flight booking, and Slack bots to help teams get started.