Analyze a complex system incident with Codex
Give Codex the incident context and logs so it can investigate, run checks and revise its findings for engineers to review.
Oct 25, 2026
Investigate this incident: [summary]. Logs are in [path]. Find the most likely root cause, the evidence for it, what you ruled out, and recommended fixes. Do not change any production system.
How to use it
1. Collect incident timeline, logs and relevant config 2. Set up Codex in a sandbox with read access to the needed data 3. Describe the incident and the questions to answer 4. Let Codex investigate and run checks 5. Ask for a findings report with evidence 6. Engineers verify the root cause before acting
Result
Codex completed the full analysis in 30 minutes. Example: Input: 3 days of service logs and a timeline. Output: a findings report naming a likely root cause with log excerpts.
Tips and gotchas
Human review: Engineers validate the analysis. Pitfall: Needs security guidelines on data and sandbox use. Privacy: Follow your security guidelines on which data and systems Codex may access. Best for: SRE, IT operations engineers (Advanced). Time saved basis: Sourced.
openai.com
https://openai.com/index/ntt-data/