Ask in Slack, answered from Confluence
Slack Knowledge Assistant over Confluence for an EdTech Company
An assistant inside Slack that answers staff questions from the company’s Confluence — routed by an agent that decides, per question, whether to search the wiki, the web, or just answer.
The problem
An education company kept its product, process and project knowledge in Confluence, and nobody could find it. Staff asked colleagues in Slack instead, the same questions were answered again every week, and new pages went unread because nothing pointed anyone at them.
What we built
An assistant that lives where people already ask: a Slack app backed by a retrieval pipeline over the Confluence spaces, kept current by watching each space’s feed for new and edited pages. It answers in the thread, from the company’s own pages, and says so when the wiki does not cover the question.
Before
- Knowledge buried in Confluence
- The same questions answered every week
- New pages go unread
How it works
Ingest and stay current. Confluence pages are loaded, cleaned and indexed; a feed watcher picks up new and edited pages so the index does not drift from the wiki.
Index for both detail and themes. The first version clustered and recursively summarised the corpus (RAPTOR) into a vector store, so broad questions met summaries and narrow ones met passages. It now runs on a knowledge graph (LightRAG) with local, global and hybrid retrieval modes.
Check the question first. A safety node screens every message before any retrieval, and blocked questions stop there.
Route by question type. An analyser labels each question — factual, analytical, conversational, real-time or calculation — and picks the retrieval mode: entity lookup, broad theme, or both.
Call only the tools it needs. A LangGraph agent chooses between the knowledge base, web search, Wikipedia, a calculator and the clock, and tags each piece of context with where it came from.
Answer with its sources. The reply is composed at temperature zero from the tagged context, cites the sources that matter, and admits a gap instead of filling it.
After
Ask in Slack, answered from Confluence
Stack
RAGKnowledge GraphLangGraphSlackWhy it holds up
An internal assistant earns trust by refusing well. Every answer carries where it came from, internal knowledge and the open web are never blended without labels, and when the wiki is silent the bot says so — which is also the signal that tells the team which page to write next.
Related service: Enterprise RAG
More case studies
Big 4 client · fully on-premise
Enterprise RAG Chatbot for a Big 4 Accounting Firm
Multimodal retrieval over a Big 4 firm’s full compliance library, tuned for citation instead of summary — every answer arrives with its source paragraph attached.
- Juniors reading thousands of pages per question
- Seniors acting as search engines
200k calls/month · 67% booking rate vs 77% human
24/7 AI Agent Call Center: 200k Calls a Month Across 60 US HVAC Brands
Voice agents orchestrating 200,000 calls a month across 60 US HVAC brands for a 5,000-person enterprise, booking 67% of calls against 77% for human agents — within 10 points of a trained team, around the clock. Built as a product, an AI front desk that answers, books and routes, not a pilot.
- Night and weekend calls unanswered
- Staffing every shift hard to justify
US$1M collected in the first 3 months
AI Payment Collection and Reminders
The companion product to the AI front desk: it chases unpaid invoices with polite, timed reminders and a payment link, and stops the moment the customer pays or replies. It collected over US$1 million for the client in its first three months.
- Chasing payment is awkward and late
- Billable time lost to collections
Get started
Talk to the engineer who will build it.
You will speak with me directly. Bring the use case, the data and the constraints, and I will give you a straight view on feasibility, effort and the first thing to ship.
