Prototype by Daniel Sedlak for the Head of Autonomous Aftersales role at Stark Future.
Not an official Stark page. The rider, bike and data are examples. The agent is live, running on Claude.
Why this case
The job is 99% of tickets handled by AI. The FAQ questions are the easy part. To get to 99% the agent has to close the tickets that go to people today, and warranty and technical cases are most of those. So this shows one of them, end to end, with nobody at Stark touching it.
What's real
The agent is live. Claude reads the bike data, decides, ships the part, books the repair and closes the ticket, using tools I wrote. The warranty decision runs as fixed rules in code, so it's auditable. Service and charging info come from Stark's public material.
What's example
Marta, her bike and its data are made up. I don't know exactly what Stark logs from the bike, so the data fields are placeholders. With access to Stark's Zendesk history, the rules and answers would come from real solved cases.
Who built it
Daniel Sedlak. 30+ years racing Motocross, Enduro and Supercross. I build and run live AI agents for customer facing businesses. Built this with Claude.
One warranty case · zero people
A battery claim, solved in one conversation
Marta's VARG EX stopped charging past 78%. Tap the first question in the chat and watch the agent read the bike data, approve the warranty, ship the battery and book the repair.
The rider's view
Signed in as Marta Ruiz
VARG EX
- Delivered
- 12 Apr 2025
- Riding hours
- 142 h
- Home
- Barcelona
Scroll down to see what Stark's team sees.
What Stark's team sees
Updates as the agent works
Ticket
Waiting for the rider
The difference
Person-first path
Triage→Dealer diagnosis→Warranty review→Parts order→Booking
Five handoffs, each one a wait for the rider.
Agent-first path
One conversation
The team gets the finished ticket and the fault data for Engineering.