What is omnidispatch?

OmniDispatch is an AI-driven system for responding to critical infrastructure failures, built as a hackathon project for the Microsoft Agents League 2026 competition. The stated problem it addresses is that when something like a power transformer overloads, a typical utility company takes around 47 minutes to get a repair crew dispatched, because operators must manually check sensor data, verify compliance rules, and find available certified workers across three separate systems. OmniDispatch aims to bring that to under three seconds by running three AI agents at the same time and requiring a human to approve the final dispatch. The three agents each handle one domain. The first analyzes the incoming sensor alert and classifies its severity. The second checks the alert against safety regulations and labor policies by searching a document index. The third looks at a database of field technicians to find someone nearby who holds the right certifications and is currently available. All three run in parallel, and a custom aggregation step combines their outputs without triggering another AI call. The dispatch is then held until a human operator approves it, at which point a signed audit record is generated. The interface lives inside Microsoft Copilot and shows an interactive map where technicians can be routed. A control room dashboard streams real-time alerts, displays a visual timeline of how long each agent took, and plays voice announcements for critical incidents. A PDF export produces a compliance report with cryptographic identifiers for each dispatch decision. The infrastructure runs on Azure: Azure AI Foundry hosts the agents, Azure OpenAI provides the language model, and Azure AI Search handles the policy document retrieval. A Node.js server exposes tools to Copilot via the MCP protocol, and the control room dashboard is built with React. Everything can be provisioned with a single Azure Developer CLI command using the included infrastructure-as-code templates. The project also includes fallback mock implementations for all the AI components, so the system can be tested without live Azure credentials. It is presented as a demonstration of the Microsoft Agent Framework's concurrent orchestration and human-approval patterns.