Your agent works in a demo, but does it hold up in production? In this episode of #AzureFriday, Scott Hanselman sits down with Sebastian Kohlmeier to explore Foundry Observability. They cover how to trace behavior, evaluate outputs, and continuously optimize agents, with a practical end-to-end workflow for moving from prototype to production. Watch the full episode: https://msft.it/6046vAqyA
Strong point on production readiness. For teams adopting AI agents in business workflows, observability should sit beside identity, data access, and change management from day one. The demo-to-production gap is where many governance issues appear.
For many, traces seem to be a fundamental building block of telemetry. Computers operate through a series of calls, with each procedure triggering the next, resulting in a cascading chain of invocations. When telemetry is organized around traces, it appears to be an intrinsic part of the machine’s architecture. Tracing essentially records the machine’s calls. While this intuitive understanding is helpful, it falls far short of providing a comprehensive grasp of systems. https://humainary.io/notes/to-observe-agents-we-need-signs-not-traces/
Can Topological Data Analysis (TDA) be implemented on the Azure platform? Here is an example of one of the analyses I conducted. https://www.linkedin.com/posts/adalberto-lopez-morales-077b011b9_tda-datascience-quantitativefinance-ugcPost-7486136880200298496-bTw6/?utm_source=share&utm_medium=member_android&rcm=ACoAADLjGxsBk8Bnh2q2gJ3OXbdEkQ5javtW-0M
865
98
This resonates a lot — most agent failures in real deployments aren't model issues, they're blind spots in tracing and evaluation once traffic gets messy. Baking observability in from day one, not bolting it on later, seems to be the real differentiator for production-grade agents. 🔍 How are teams balancing trace granularity with latency overhead at scale?