Agentic AI-Powered BI

Executives need to explore business information without relying on a new analyst-built dashboard for every question. Treating a chat interface as the product would lose the rigour analytics demands.

I worked across product, design and engineering on PRDs and product/design operations deliverables. The focus was a shared model of the product, its data boundaries and the workflows that agents would support.

I defined five core modules and collaborated on streaming ingestion, row-level security and a microservices approach. Product requirements and interaction design connected AI capability to a usable executive experience.

Conversation is an entry point, not the whole experience. Trustworthy analytics still needs clear data boundaries and a way for users to inspect, compare and act on results.

Five core modules with agentic workflows, streaming ingestion and row-level data access in product planning — conversation sits alongside inspection and direct manipulation rather than hiding the product behind a prompt.