Just Now Possible When AI Becomes Your SRE: How Incident.io Is Automating Incident Response
12 snips
Nov 6, 2025 
Join Lawrence Jones and Ed Dean from Incident.io as they dive into the revolutionary impact of AI on incident response. Lawrence, a founding engineer leading the AI team, and Ed, the product lead for AI, share insights on how AI compresses investigation times from hours to mere minutes. They discuss the importance of user trust in AI, simplifying complex systems, and using past incidents to improve future responses. Explore their innovative Slack integration and the fascinating concept of AI as a company's immune system in action!
AI Snips
Chapters
Transcript
Episode notes
Blend Retrieval Methods, Favor Deterministic Tags
- Don't rely solely on embeddings; combine text similarity, summaries, and re-ranking for reliable retrieval.
- Use deterministic tags and re-rank results with frontier models to improve precision and debuggability.
Re-Rank To Turn Fuzzy Matches Into Useful Results
- Re-ranking vector search results with an LLM produces higher-quality shortlists than raw vector scores.
- Combining deterministic pre-filtering (tags) with LLM re-ranking yields both debuggability and accuracy.
Prioritize Concise Slack Reports, Hide Deep Reasoning
- Post concise incident summaries in Slack and keep detailed reasoning in threads or a dashboard for deep dives.
- Favor signal over noise and surface only high-confidence claims in the main channel to avoid distraction.