Run AI SREs without burning token budgets | ODSP928

This session from Microsoft Build 2026 addresses the $2 per alert cost barrier preventing enterprises from deploying AI SREs as first-line triage solutions. It provides mathematical cost breakdowns for large enterprises, compares cheaper models like DeepSeek against Opus with accuracy trade-offs, and explains practical optimizations including LLM-native grouping over deterministic rules and cached context window reuse. Designed for SREs, platform engineers, and technical leaders managing enterprise alerting systems who need to reduce AI investigation costs.