Agent Evaluation Series | Governance, Lifecycle Gates, Operating Agents | FastTrack TechTalk
In this final part of the FastTrack TechTalk series, Microsoft architects explore the governance, evaluation, and operation of AI agents in production settings. The session highlights that traditional testing methods don't suit AI agents, advocating for continuous evaluation and lifecycle gates to maintain reliability. Key topics include the unique governance needs of AI agents compared to traditional D365 solutions, early detection of agent quality drift, and covering the entire evaluation lifecycle from design to operation. The talk emphasizes using evaluation gates for informed decision-making, employing production strategies like shadow mode and A/B testing, and underscores the importance of monitoring and observability. It also discusses creating a continuous feedback loop to boost agent performance.
Video 3w
Login now to access my digest by 365.Training
Learn how my digest works
Features
- Articles, blogs, podcasts, training, and videos
- Quick read TL;DRs for each item
- Advanced filtering to prioritize what you care about
- Quick views to isolate what you are looking for right now
- Save your favorite items
- Share your favorites
- Snooze items you want to revisit when you have more time