Keeping community human while scaling with agents
Not a pilot. Vercel deployed AI agents across 281 community threads—and their non-engineer community manager built it herself using Claude and plain English prompts.
Why it matters
Vercel's 'Community Guardian' demonstrates practical agent deployment at scale: automating triage and routing while preserving human expertise for high-touch moments. The story matters because it shows how non-engineers can ship production agents using existing tools (Claude, AI SDK, Vercel Workflows), lowering the barrier to agentic workflows beyond research labs.
The key facts
9 to know281 unique users helped in first 23 days
4,716 first responses triaged and context-gathered by agents
1 in 8 'ghosted' threads revived, resulting in 23 confirmed solutions
1,400+ agent runs in two-week period
4 duplicate threads detected and auto-closed at 95%+ confidence
Agent checks community every 10 minutes using Vercel Workflows (durable execution)
Agents handle routing, triage, follow-ups; humans handle complex pair-debugging and relationship building
Built by non-engineer community manager using Claude (via AI SDK) + AI Gateway + Vercel Workflows
Used multi-model approach (ability to switch models without API key changes via AI Gateway)
Go to the source
Vercel Blogvercel.com
Publisher excerpt: At Vercel, our developer community is at the heart of everything we do. It's how we stay closest to the people using what we build. As our community grew, automation helped us scale. But questions still got lost, routing took time, and context switching pulled us away from the work that actually…