To integrate the new app assistant ahead of the 2025 Combine, the NFL worked with longtime Microsoft development collaborator Southworks to integrate a conversational AI assistant into the existing Combine platform. The architecture brought together
Azure OpenAI for natural language processing;
Azure Cosmos DB for fast, structured data access; and
Azure Container Apps for scalable, event-day performance without infrastructure management. As a result, the assistant could interpret sideline queries, analyze and compare player data, and respond in real time, providing scouts with accurate player data and fast performance.
At the core of the assistant is AzureOpenAI Service, which runs the latest GPT-series model. To power the assistant’s experience, developers built on a tightly integrated stack of Azure services that handle natural language interaction, data processing, and scale.
“Now I just ask a question as I would ask you a question—and get what I need in seconds. It used to take hours, sometimes not even possible until after the Combine,” says Jacqueline Davidson, VP ofFootball Research for the Tampa Bay Buccaneers.” For Davidson, that shift turned a time-consuming process into a real-time edge. The solution didn’t just speed things up; it brought insight exactly when she needed it.
Supporting the assistant’s intelligence isContainer Apps, hosting both the assistant’s API and the background services responsible for fetching and normalizing data from the NFS’s systems. ChoosingContainer Apps gave the team a fully managed compute layer that not only scaled on demand for event-day traffic, but supported enterprise-grade reliability, security, and compliance—all without the overhead of managing infrastructure.
At the data layer, Azure Cosmos DB serves as the data store behind the assistant, caching structured player information so the assistant can respond quickly. In future iterations, Azure Cosmos DB will also facilitate multi-turn interactions by acting as the assistant’s memory layer, but in this release, all responses are single-turn and stateless.Together, this architecture gave the Combine’s new AI assistant a production-grade foundation—highly secure, scalable, and responsive—ready for real-time sideline use within the app that teams already trusted.
For club personnel, the benefit was clear:natural language queries turned manual reports and comparisons into instant answers during critical Combine moments.