Ep216: Powering AI-enabled Operational Insights with Amazon Bedrock

From alert to root cause in one minute - how PagerDuty built autonomous incident response on Amazon Bedrock, and the future of triage and trust.

Topics Include:

• PagerDuty's agents must perform during 2am outages — stakes are high
• Software shipping accelerated dramatically; production environments largely did not
• A 9:30pm slowdown traced to a race condition solved two years earlier
• The fix was documented — but the context wasn't at hand
• PagerDuty Advance ships four agents: SRE, Scribe, Shift, Insights
• Why four, not one? Focus and predictability in non-deterministic systems
• Saurabh Shanbhag: Bedrock is far more than a model service
• Zero data retention, PrivateLink, TLS — why enterprises pick Bedrock
• Frontier models everywhere burns tokens; classify, route, distill, fine-tune
• SRE agent triages alerts before you even join the call
• One minute to root cause — context beat raw intelligence
• Human surfaces versus machine surfaces: MCP and CLI move fastest
• "The model eats the harness" — every upgrade invalidates foundational components
• Feeding agents everything failed; compartmentalised investigation threads work better
• New York Life's three stages of trust, and the seatbelt override that wasn't

Participants:

• Tom Hogarty - Senior Director Product Management, PagerDuty
• Saurabh Shanbhag – Sr Partner Solution Architect, Amazon Web Services  

See how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon.com/isv/