A RAG assistant that cut support tickets by a third

Illustrative scenario — a representative example of the kind of AI development engagement this describes, not a specific named client.

A B2B SaaS support team is drowning in repetitive questions. The answers all exist — scattered across a 400-page docs site, a Notion runbook, and three years of resolved Zendesk tickets — but nobody can find them fast enough.

Challenge

The knowledge isn't missing; it's unsearchable. New hires spend weeks learning where things live, and customers wait hours for answers a senior engineer could give in thirty seconds. Off-the-shelf chatbots hallucinate confidently, which for a developer-facing product is worse than no answer at all.

A team in this position needs something that will:

• Answer from their content, with citations, and say "I don't know" when it doesn't know
• Stay current as docs change daily
• Run inside their own infrastructure for data-residency reasons

What we build

An internal retrieval-augmented (RAG) assistant embedded in the existing admin app.

• A Python ingestion pipeline that chunks docs, runbooks, and resolved tickets, embeds them with OpenAI, and stores vectors in pgvector alongside the existing Postgres
• Hybrid retrieval (semantic + keyword) with a re-ranking pass, so exact error strings and conceptual questions both work
• A Next.js chat surface with inline source citations and a one-click "escalate to human" handoff
• Nightly re-indexing wired to the docs CI, so the assistant never drifts from the published source

Every answer links back to the passages it used. When confidence is low, it routes to a human instead of guessing — the citations are what earn a support team's trust.

Results

Figures below are illustrative of the outcome this kind of work targets — not a measured result from a named client.

• −34% inbound support tickets, driven by in-app self-service deflection
• First-response time on escalations down from hours to minutes
• New-hire ramp shortened, as the assistant becomes the team's institutional memory

The assistant handles the long tail of "where is this documented" questions, so engineers spend their time on the problems that actually need a human. This is exactly the kind of internal tool our AI development work is built to ship.