Case study · AI / ML
A RAG-based assistant in the help centre and support inbox that answers from docs and past tickets, and hands off to humans with context.
- Client
- A B2B software company with a global customer base
- Industry
- B2B SaaS
- Region
- North America
- Duration
- 8 weeks
- Team
- 2 AI engineers, 1 full stack engineer
Ticket volume grew faster than the support team, first response times slipped past a day, and most questions were already answered somewhere in the docs.
42%
of tickets resolved without an agent
−65%
first response time
93%
answer accuracy on the eval set
Step 001
Eval set first
Built 300 real questions with approved answers to measure quality before and after every change.
Step 002
Retrieval
Indexed docs, release notes and resolved tickets with hybrid search and re-ranking in pgvector.
Step 003
Assistant
Claude-powered answers with citations, confidence thresholds and clean human handoff.
Step 004
Operate
Tracing in Langfuse, weekly eval runs and a dashboard for deflection and cost.
- Claude
- Python
- FastAPI
- pgvector
- LangChain
- Langfuse
- Next.js
- Sentry
Tell us about it and we'll reply with how we'd approach it, usually within one business day.