AI for Coding: Draft an Incident Postmortem From Logs and Chat

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-ai-coding-incident-postmortem-draft-r8a1-creators

1. What is the primary advantage of using AI when constructing an incident timeline from multiple sources?

   1. AI can decide which corrective actions the organization must implement
   2. AI can determine which team members should be blamed for the incident
   3. AI can automatically assign performance ratings to involved engineers
   4. AI can merge heterogeneous sources like logs, alerts, and chat transcripts into a unified timeline

2. After AI generates a postmortem draft that identifies 'Sarah deployed the faulty configuration,' what is the appropriate next step before sharing the document internally?

   1. Add additional details about Sarah's technical background and experience level
   2. Submit the draft exactly as written since AI accurately documented what happened
   3. Delete the mention of Sarah entirely since AI should never name individuals
   4. Replace the personal reference with a focus on the system conditions that allowed the deployment to proceed

3. Which of the following tasks falls within AI's demonstrated capabilities when assisting with incident postmortems?

   1. Identifying decision points and who acted during an incident
   2. Deciding which team members should receive formal warnings
   3. Determining staffing changes needed to prevent future incidents
   4. Judging whether an engineer's performance met expectations

4. An AI tool analyzing incident logs suggests that 'inadequate testing coverage' is a root cause. Why must a human review this suggestion before finalizing the postmortem?

   1. The suggestion might conflate a contributing factor with the root cause and requires judgment about systemic issues
   2. Testing coverage cannot actually affect incident outcomes
   3. AI always produces correct conclusions about testing processes
   4. Human review is unnecessary since AI analyzed all available data

5. When an AI-generated timeline has gaps showing 'evidence missing for 14:00-14:15,' what does this indicate about the postmortem process?

   1. The incident was not serious enough to warrant investigation
   2. The AI failed and should be discarded
   3. All incidents should have complete documentation
   4. The timeline accurately reflects where log data or documentation is incomplete

6. A team lead asks AI to identify which developer 'caused' a production outage. What is the fundamental limitation of this request?

   1. AI does not have access to the necessary source code
   2. AI cannot assign blame or judge individual performance—only humans can make accountability decisions
   3. AI cannot read log files from production systems
   4. AI lacks access to performance reviews of the developers

7. You receive an AI-generated postmortem that begins with 'The outage occurred because of a misconfiguration.' Why might a human want to revise this opening?

   1. The statement attributes causation without explaining the systemic conditions that allowed the misconfiguration to be deployed
   2. The statement is too specific and should be removed entirely
   3. AI is never allowed to describe what caused incidents
   4. Misconfigurations cannot cause outages

8. In a blameless postmortem culture, how should the sentence 'Tom ignored the warning alert' be rewritten?

   1. The alert system sent a warning that was not acted upon before the failure occurred
   2. Tom should have responded to the warning but chose not to
   3. Tom was the only engineer available and could not respond in time
   4. The warning system failed to escalate properly

9. Which organizational element can AI NOT reliably detect from analyzing incident logs and transcripts alone?

   1. Time gaps between related system events
   2. The sequence of technical events leading to failure
   3. Team culture issues or staffing constraints that may have contributed
   4. Changes in system configuration during the incident window

10. When AI produces a customer-facing summary of an incident, what additional consideration should a human apply before publishing?

    1. Replace the summary with a technical deep-dive
    2. Ensure the language appropriately represents the incident to external stakeholders without exposing internal system details or assigning blame
    3. Remove all technical details since customers do not understand technology
    4. Add blame statements to show accountability

11. What input materials would you provide to AI to help it construct the most comprehensive incident timeline?

    1. Only the final incident report from the previous year
    2. A list of engineers who were online during the incident
    3. PagerDuty alerts, Slack incident channel transcript, and deployment logs
    4. A summary email from the incident manager

12. An AI-generated postmortem lists five action items. Who should determine the priority and scope of these items?

    1. The AI should be reprogrammed to only suggest one action item
    2. No one—they should all be implemented immediately
    3. The AI system that generated them, based on impact analysis
    4. The engineering team, through consensus and organizational decision-making

13. A postmortem AI tool suggests that 'inadequate staffing on weekends' contributed to an incident. Why should humans evaluate this suggestion rather than accepting it directly?

    1. The suggestion requires human judgment about whether it reflects a real organizational issue or an artifact in the data
    2. AI never suggests staffing-related factors
    3. Staffing has no relationship to incident outcomes
    4. Human evaluation is unnecessary for staffing suggestions

14. What makes a postmortem 'blameless' rather than simply 'accurate'?

    1. It includes no recommendations for improvement
    2. It is written by AI rather than humans
    3. It does not mention any technical details
    4. It focuses on what went wrong without assigning fault to individuals

15. When an AI timeline shows that multiple alerts were triggered but no human responded for 20 minutes, what question should the postmortem investigation explore?

    1. Why the alert system was designed to require any human response at all
    2. Whether the alert thresholds were set too sensitively
    3. Who was specifically responsible for ignoring the alerts
    4. How to punish the on-call engineer