The Bottleneck
Management and functional teams (starting with Internal Control) were overwhelmed by static PowerBI dashboards. While the data existed, extracting actionable insights required users to log in, filter views, and manually interpret trends. This "passive" consumption model slowed down decision-making.
The Logic Flow
- Step 1: The Trigger Power Automate → Connects directly to the PowerBI Semantic Model via DAX queries to extract the latest compliance and performance data.
- Step 2: The Intelligence → The raw JSON data is passed to AI Hub (LLM) within the secure L'Oréal tenant.
- Step 3: The Delivery → The LLM generates a structured HTML summary, which is automatically emailed to the distribution list via Outlook
The "Secret Sauce"
System Prompt Engineering The technical challenge wasn't the connection; it was the hallucination risk. I led the team to master System Prompt Engineering to ensure the AI acted as a strict analyst, not a creative writer.
Logic: We designed prompts that enforced strict hierarchy (e.g., separating "Zone" vs. "Country" logic) and forced the output into a specific HTML format for executive readability.
The ROI & Roadmap
- Validation: Successfully piloted with the Internal Control team, automating their static analysis reporting.
- Scalability: Currently expanding the architecture to the CMI (Consumer Market Intelligence) team to automate their market share newsletters.
- Innovation: Validated the first internal use case of Generative AI for corporate reporting within the Zone's strict compliance framework.