Project Summary
Finance stakeholders needed instant, plain-English answers to budget variance questions without waiting for analyst reports or manual data pulls from disparate systems. We built a custom Copilot Studio bot grounded in live SharePoint policy definitions and SQL budget vs. actuals data, enforcing business rules and role-based category masking, and triggering Power Automate workflows when variance explanations, approval tasks, or ownership notifications were required — transforming BI insights into tracked, accountable operational actions within the existing Power Platform ecosystem.
Challenge
Every budget variance question required analyst mediation, creating decision bottlenecks, delayed action cycles, and no self-service path for finance leaders — while sensitive financial data categories needed to remain masked based on the requestor's authorization level at query time.
Key Outcomes
- Natural language budget Q&A grounded in live SharePoint policy definitions and SQL actuals data — no analyst mediation required for standard queries
- Role-based security masking sensitive financial categories and limiting department visibility to authorized scopes only at the query layer
- Power Automate workflow triggers creating approval tasks, requesting formal variance explanations, and notifying department owners in Teams
- Guided prompt experiences for the most common finance workflows: variance explanation requests, top driver analysis, and trend change investigation
- Fallback behaviors when required data inputs are missing — the Copilot asks for minimum needed context rather than failing silently
- Full interaction logging, audit trail for all actions, and delivered admin documentation including a prompt pack for leaders, finance analysts, and department owners