AI SRE Summit 2026
as the organisers put it
Join the Alliance to learn from enterprise teams already using AI SRE in production to automate incident response, and eliminate toil.
| title | author | length |
|---|---|---|
| AI IN SRE: HYPE VS REALITY IN 2026 | Stefana Muller, Charity Majors, Sharone Revah Zitzman, Itiel Shwartz | 50m |
Panelists report concrete wins and limits: auto-generating postmortems, Slack-thread TLDRs and documentation-from-code cut toil and drop MTTD/MTTR; narrow… | ||
| HOW TO BUILD YOUR OWN AI AGENT: THE COMPLETE ARCHITECTURE FOR ENGINEERING TEAMS | Viktor Farcic | 22m |
Architectural blueprint for a production agent that ties system context, tool descriptions, and just-in-time knowledge via an ingestion pipeline from git… | ||
| The Cat and Mouse Game of Infra Cost vs. Latency | Parakh Jaggi | 29m |
Modern AI inference shows P50 ~180ms, P95 ~850ms and P99 ~1.5s so tail events dominate sessions; only ~20% of wall time is GPU compute and ~80% is overhead… | ||
| DO WE STILL NEED TO "OBSERVE" THE FUTURE OF AI & OBSERVABILITY | Guy Menahem | 28m |
Claims that AI eliminates no PromQL by translating plain English into vendor queries, automates correlation and causal-finding, and runs SRE agents that draft… | ||
| AGENTS NEED GROUND TRUTH: WHY IAC IS THE FOUNDATION OF AUTONOMOUS OPS | Eran Bibi | 22m |
Argues that Infrastructure-as-Code must be the agent's source of truth, or autonomous fixes create silent configuration drift; shows three real failure modes… | ||
| THE SANITY CHECK: OBSERVABILITY FOR AI DATA PIPELINES | David Aronchick | 22m |
A chatbot answers “365 days” but the true policy is 30 days because a rogue Salesforce feed pollutes embeddings; the pipeline is a black box so top-line… | ||
| YOUR AI DOESN'T KNOW WHAT THINGS COST | Corey Quinn | 25m |
A cautionary pattern: proofs of concept use big models and cheap staging but ship unchanged to production, producing 200,000 tokens per diagnosis at $1–$5 each… | ||
| YOU CAN'T AI YOUR WAY OUT OF A BROKEN PLATFORM | Brittany Woods | 22m |
Claim: layering AI atop a fragmented platform multiplies failures instead of fixing them. Example: asking AI to deploy a service still needs repo discovery… | ||
| STOP TREATING AI SRE AS JUST AN ENGINEERING PROBLEM | Blake Sherwood | 32m |
Smarsh converts SRE signals into planning intelligence by deploying agent classes on assistive, workpack, autonomous layers that surface capacity, dependency… | ||
| CONTEXT ENGINEERING FOR SELF-HEALING AI SRE | Asaf Savich | 43m |
Komodor analyzes nearly 1.8M incidents and reduces Kubernetes chaos by moving from brittle deterministic runbooks to an AI agentic system named Claudia agent… | ||
| Your AI Agent Has No SLO | Andrew Espira | 22m |
Built an agent observability stack that wires agent traces and attributes into OpenTelemetry → Prometheus → Grafana, defines SLOs and error budgets, and alerts… | ||
| If AI Writes the Code, Who Runs Production? | Alan Shimel | 19m |
Claims that 41% of code is generated by AI and development velocity increases 2x–4x, with tools like GitHub Copilot and Claude accelerating releases. Incidents… | ||
| nothing matches | ||