Internal Knowledge Assistant
The company's first GenAI product: an internal copilot grounded on policies, SOPs and training documents. Employees ask, it answers with the right source, in seconds instead of a hunt.
Role
Product Owner & Data Scientist
Timeline
2025 · 3 months end to end
The challenge
Company knowledge lived across SharePoint, policy documents, training material, process guides. Finding anything meant knowing where to look or whom to ask, and most people knew neither. Time quietly leaked everywhere.
The decision
Chose Copilot Studio inside M365 over Standalone chat app
The company already lived in Teams and SharePoint. Meeting people in their existing tools beats asking them to adopt a new one, and the fastest path to a first AI win builds trust for everything after.
How I approached it
- 01Started with use-case collection across departments; instant access to internal knowledge surfaced as the most universal pain and the lowest-risk first bet.
- 02Scoped it deliberately small in the PRD: public internal SharePoint knowledge only, no sensitive data, so governance stayed light and shipping stayed fast.
- 03Grounded the copilot on policies, SOPs and training docs so every answer points to its source.
- 04Shipped in three months and used its adoption as the proof that funded the bigger AI portfolio.
Impact
- ✦'Ask around and wait' became a ten-second conversation.
- ✦Answers arrive with the source document, not just the claim.
- ✦First GenAI product in production, the trust-builder for the portfolio.
- ✦Zero new tools to learn: it lives where employees already work.
Stack
- Copilot Studio
- Microsoft 365
- SharePoint
- Azure AD