Currently I am tuning up our Entra ID logs with detections and FP rules. I find that each session there is a lot of tokens burned up on understanding the shape of the log data. What can I do to help it understand the important fields faster. Is that an “SOP” for the organization or would I need to build a “skill” around Entra ID logs?
Hello, I think you should be able to use the new memory feature we introduced recently: AI Memory for AI Agents and AI Sessions
You’ll need to enable the ai_memory.* permissions, but then you can either use a user AI session to create a synthetic memory for the target agent, or you could tweak the prompt of the target agent to tell it to create memories when it has trouble figuring something out to record it so it’s faster next time.
Let me know how it goes! 
1 Like