AI assistant deflects 60% of support tickets for Quantia
−60% tickets needing humans
Challenge
Quantia runs a fast-growing commerce marketplace. Support was drowning: tickets tripled in six months, average response time hit 8 hours, and customers complained.
Approach
Rather than replacing humans, we designed a co-pilot assistant: answering repeat questions from docs, handing complex cases to people. We started with quality evals and guardrails before writing any code.
Solution
We built a RAG system over Quantia’s knowledge base and historical tickets, with:
- Automatic document extraction and sync
- Guardrails: no fabrication, cite sources
- Human handoff on low confidence
Results
−60%
tickets needing humans
2 phút
avg response time
94%
CSAT score
4 tuần
to production
“Their AI team shipped an LLM-powered support assistant that now deflects most incoming tickets. Senior engineers, no fluff, real results, on budget.”
Tech stack
PythonLangChainPineconeGPT-4oFastAPI
Start a similar project
Tell us about your challenge — we'll bring the team.
