Case study · AI Automation · Live

Internal knowledge assistant & ticket triage

RAG over internal docs plus auto-triage on inbound tickets, with confident hand-off to humans on ambiguity.

Industry
B2B SaaS
Duration
6 weeks
Outcome
First response time: 6h → 38m
Status
Live
StackNext.jsPostgres + pgvectorAnthropic ClaudeInngestLinear API
The challenge

What we walked into.

A growing support team was drowning in tickets that mostly had answers buried in docs and old conversations. Hiring more agents was an option; cutting time-to-answer was a better one.

The approach

How we built it.

  1. 01RAG over docs, runbooks and resolved tickets — not a generic chatbot, a team-specific assistant.
  2. 02Auto-classification on inbound tickets with confidence threshold and human hand-off.
  3. 03Suggested replies in the agent UI rather than full autonomous responses.
  4. 04Feedback loop: every accepted reply improves the retrieval ranking.
Outcomes

Numbers from the real rollout.

Avg. first response
6h → 38m
Auto-classified
73%
Agent CSAT
+18%
Coverage
9 doc sources
More work

Other things we’ve built.

Let's build

Have a system in mind? Let's sketch it together.

Drop your email and a line about the problem. We'll reply within one working day.

Or write directly to hello@corefluxsolutions.com — we read everything.