Vercel
Automating 90% of support tickets: How we built Vercel's support agent — Ship 26 NYC
An in-house support agent, built by two engineers in three weeks, now resolves 91% of Vercel's tickets autonomously.
KM
Kathryn Middleton
Engineering Manager, Agent Experiences · Vercel
18:15
The numbers, across the talk3 moments · 18:15
01 · Autonomous customer support
Vercel
Vercel replaced an off-the-shelf support vendor with an agent built on its own platform (AI SDK, AI Gateway, Vercel MCP, Marketplace search). It is now the first line of defense on every incoming ticket, resolving the large majority without a human.
00:01:03
▲91%
Quality
Tickets auto-resolved
“we completely transformed our support operation to resolve now 91% of tickets, completely autonomously”
00:01:03
▼3 weeks
Cycle time
Time to production
“how we shipped to production in just three weeks. And that was seven months ago.”
00:02:02
●7,000/wk
Scale
Tickets resolved last week
“Just last week, it resolved over 7,000 tickets.”
00:02:02
●130,000
Scale
Cases handled to date
“in total, it handled over 130,000 cases fully autonomously”
00:02:02
▲5,000 hrs/mo
Productivity
Engineer hours saved
“This has saved our support engineers over 5,000 hours every month.”
00:02:02
▲= 31 engineers
FTE savings
Equivalent headcount
“that's equivalent of adding 31 additional support engineers to our team”
00:16:16
▲+20 pts
Quality
Above industry resolution
“It outperforms the industry by 20 points.”
00:16:16
●2 engineers
Other
Team that built it
“We built the support agent in three weeks with the team of two using features [available] to every Vercel customer today.”