STOP TREATING AI SRE AS JUST AN ENGINEERING PROBLEM
summary
Smarsh converts SRE signals into planning intelligence by deploying agent classes on assistive, workpack, autonomous layers that surface capacity, dependency and risk signals back into PRDs. A prototype agent in Claude detects a predicted 4x traffic / 4% margin hit, flags a conflicting rollout and recommends scaling or schedule changes. Systems include agent core on AWS Lambda, Kubernetes metrics, and catalog analysis over an environment with ~41 TB logs/month and ~20,000 services; the prototype registers risks and improves roadmap decisions.
AI-written summary · about
chapters (9)
- 0:00Introduction and background
- 3:19Disconnect in AI SRE approach
- 6:51Classifying AI SRE agents
- 10:21Continuous validation in SDLC
- 13:39Business risk signal integration
- 17:10Agent-based risk assessment example
- 23:40Go-to-market recommendations
- 27:12AI SRE as business utility
- 30:21Future of AI SRE transformation
related talks
40mWhen Incidents Fix Themselves: AI SRE in action
44mAgentic Observability — The Path to AI Co-SRE
15mWho Owns Your SRE Stack? The Rise of AI Agents
30mArchitecting Integrity: Responsible Software Evolution in the AI Era
21mSRE, what is it all about?!
18mDeploying Humans at the Edge of SRE
29mLeadership in AI-assisted engineering
19marchive – Shane Murray