Evidence to Decision: building an AI-envisioning workshop tool on Azure

This article details the design and implementation of 'Evidence to Decision,' a custom AI workshop tool built on Azure. It addresses the common problem of lost insights from sticky notes by creating a governed, evidence-first pipeline. The tool captures participant statements, uses live voting and clustering on a shared mural, and employs Azure AI agents to draft problem framings based solely on evidence. Human reviewers approve or block these drafts, ensuring AI remains advisory. The architecture uses clean architecture with Azure Cosmos DB for event-sourced state, Azure SignalR for real-time collaboration, and Azure Container Apps with Static Web Apps for deployment. It emphasizes managed identity, Entra ID authentication, and a strict separation between canonical state and derived recommendations.