Trustlane
33 behavioural scenarios in the evaluation suite
- Independent build
- Voice agents
AI-powered EV warranty coordination that catches missing evidence before a claim is submitted.
Project information
- Independent build
- AI agents, ElevenLabs, Voice workflows, Evaluation suites
- 33, behavioural scenarios in the evaluation suite
The problem
EV warranty claims stall when evidence is incomplete. Dealers and manufacturers go back and forth with repeated information requests, and technicians are pulled off the workshop floor to answer them.
What I built
A platform that identifies missing documentation before a claim is submitted.
- Two-sided agents. Dealer and manufacturer AI agents, each working from its own side of the claim.
- Deterministic checks. Requirement checks that do not depend on model judgement, so gaps are flagged consistently.
- Voice evidence collection. ElevenLabs voice workflows call technicians to collect evidence and keep auditable transcripts.
- Humans stay in charge. Submissions are human-reviewed, and the system makes approval recommendations rather than decisions.
How it was validated
- Three consecutive internal end-to-end walkthroughs.
- 8/8 agent-turn checks and 9/9 technician-call checks in every walkthrough.
- A 33-scenario behavioural evaluation suite behind it.
33behavioural scenarios in the evaluation suite