Client case study · Customer support AI
Support AI That Resolves Tickets: Ground-Up Rebuild
Rebuilt software that answers support tickets for hundreds of customers, resolves the routine ones itself and hands the rest to a person: 500+ pull requests in about eight weeks.
The problem
Support teams wanted AI that resolves tickets end to end, not a chatbot that deflects them.
What was built
A ground-up rebuild of the production platform for an AI SaaS startup, $1M+ ARR: the agent builder, workflow automation and finance integrations. I authored 500+ pull requests in about eight weeks and reviewed, fixed and deployed about 300 more from the team.
Results
- Agents resolve routine tickets end to end for hundreds of customers and hand the rest to a person.
- Changes to customer data wait for human approval, and personal data is masked before any model sees it.
- Every release is gated by replayed-conversation regression tests and tracing.
- An event-driven backbone of 23 job queues across 12 services, watched by 34 production monitors with runbooks.