Case study / 2025
Company Policy Bot
Internal employee support experience for a policy-heavy operations environment.
01 / Context
The problem
Employees needed self-service access to policy documents without creating a dependency on HR or operations teams for repeat questions.
A support-heavy internal team had policy answers spread across documents, updates, and repeated employee questions.
02 / Contribution
What I built
Built a Copilot Studio agent grounded in approved knowledge sources, with topic routing, response guardrails, and escalation paths.
The design decision
The useful path was not to create a generic chatbot, but to design a guided policy assistant that knew where its answers came from.
Scope
- Employee self-service for policy, SOP, FAQ, training, form, and departmental knowledge
- Access through Microsoft Teams and SharePoint-facing channels
- Grounded answers from approved SharePoint document libraries
03 / Structure
How it works
Designed as an enterprise knowledge assistant that lets employees ask natural-language questions across trusted organizational documents.
- Copilot Studio handles conversational orchestration, topic handling, context, generative answers, and safety controls
- SharePoint acts as the source of truth for indexed policy and departmental documents
- Microsoft Entra ID and SharePoint permissions keep access aligned with organizational roles
Through the workflow
- Employee opens the assistant in Teams or SharePoint
- Copilot Studio interprets the query and retrieves relevant SharePoint knowledge
- The response is generated from authoritative content and returned in the conversation
04 / Operation
Controls and upkeep
- Single sign-on with organizational accounts
- SharePoint permission trimming
- Development, UAT, and production environment path
Keeping it useful
The knowledge lifecycle stays manageable because updated SharePoint documents can be re-indexed and reused by the assistant.
05 / Result
The outcome
Provided a safer first stop for knowledge-heavy policy support using approved organizational content.
Employees gained a safer first stop for policy questions while the support team kept control over source content and escalation.
Discuss a similar project