60% deflection on tier-1 support tickets without losing CSAT.
A support chatbot trained on Notion docs + Intercom history that handles password resets, billing questions, and onboarding — escalating only the 40% that need a human.
- 60%tier-1 ticket deflection
- 4.6 / 5CSAT (vs 4.4 with humans only)
- 3.1×tickets resolved per support hour
- $0.04average cost per resolved conversation
The starting point.
Support team of 4 was being crushed by repetitive tier-1 tickets — password resets, billing questions, plan-comparison questions — leaving no time for deeper customer success work. Previous attempt at a chatbot from a SaaS vendor had a CSAT of 2.1/5 and was turned off after 6 weeks.
How we got there.
- 01Audited 90 days of Intercom conversations, categorized into 23 ticket types, identified the 12 high-volume / low-complexity types as automation targets.
- 02Built a RAG pipeline pulling from Notion (product docs), Intercom (resolved tickets), and Stripe (live account state).
- 03Designed escalation flow with full conversation handoff — the human picks up where the bot left off, with no 'tell us your problem again'.
- 04Shadow-tested for 10 days alongside human responders before going live.
What we shipped.
A custom React chat widget on app.[client].com powered by Claude with retrieval over the client's docs and ticket history. Handles tier-1 questions in <2s, escalates with full context, and the team can see every bot conversation in Intercom alongside human ones.
- Anthropic Claude
- Pinecone
- Intercom
- Notion API
- Custom React widget
“[QUOTE PLACEHOLDER — Get a quote from the head of support about what the team does with their reclaimed hours.]”
Send us your most expensive operation.
We'll have an audit on your desk in five days.
One PDF. No deck. No obligation. We'll tell you whether AI is the right answer for it — and if it is, we'll quote the build the same week.