The Hard Part isn't the Model Call

Join us in person at Farm Credit Services of America (5015 S 118th St, Omaha, NE 68137 "\https://goo.gl/maps/k6bW7GGA5zSduiLA8\")). Our sponsor provided food and drinks 15-20 minutes before the start of the meeting. The Hard Part isn't the Model Call \| Lessons from putting Azure AI Foundry into production This discussion posits that the hardest part of enterprise AI isn't the model call itself, it's the surrounding infrastructure: governance, identity, networking, cost attribution, and observability. Presented by the AI Engineering team who operate as a Center of Excellence, we will discuss how routing Azure AI Foundry behind an API Management gateway solves problems like fragmented auth, scattered cost tracking, and inconsistent logging. Three real case studies include the credit narrative generation, document metadata validation, and agentic fraud detection, illustrate the friction points teams hit going from POC to production. The core message is that data residency, identity, and deployment decisions must be locked in before code is written, and telemetry must be mandatory on both the gateway and app side. It closes with three actionable takeaways any team can apply immediately. The speakers are Shane Fitch and Dave Karas Shane Fitch Avid AI Enthusiast with 11 years at Farm Credit Services of America. Most of his career has been spent as an application developer and then a lead application developer. He developed a passion for AI with the first public release of Chat GPT and has been learning about and implementing AI powered solutions since then. Dave Karas is an AI Engineer at Farm Credit Services of America, where he stepped into the role about six months ago after more than five years with the organization. Before moving into AI full-time, he was a developer on the Digital Customer Experience team, where he was…