tubeshowcase●1,669,063 videos♡ My collection 0
← Learn to Code

Learn to Code09:15

Run AI SREs without burning token budgets | ODSP928

Microsoft Developer ↗

From the creator

Most AI SREs have an underlying cost per investigation of about $2 per alert. This makes it difficult for most enterprises to use AI SREs as a first-line solution for triaging alerts, based on token cost. In this session, we review the math and optimizations and show practival ways to bring the LLM cost of AI SREs into a manageable range, where running them on every alert becomes viable. 𝗦𝗽𝗲𝗮𝗸𝗲𝗿𝘀: * Natan Yellin 𝗦𝗲𝘀𝘀𝗶𝗼𝗻 𝗜𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻: This is one of many sessions from the Microsoft Build 2026 event. View even more sessions on-demand and learn about Microsoft Build at https://build.microsoft.com ODSP928 | English (US) | Developer tools & frameworks Pre-recorded | (300) Advanced #MSBuild Chapters: 0:00 - Explaining naive cost model for large enterprises with alert volumes 00:00:54 - Breakdown of estimated annual cost of $730,000 per Fortune 500 company 00:03:30 - Using Cheaper Models as a Potential Optimization 00:03:37 - Comparison Between Opus and DeepSeek Models 00:03:55 - Accuracy Trade-offs with Lower-Cost Models 00:04:47 - Discussion of Model Cost Optimization 00:06:30 - Adopting LLM-Native Grouping Over Deterministic Rules 00:07:39 - Reusing Cached Context Windows to Reduce Cost 00:08:46 - Benefits of Running AI Investigations on All Alerts

Published 2026-06-03 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.

Keep exploring Learn to Code