Failure Mode Analysis: Using AI to Identify Emergency Plan Vulnerabilities

Failure Mode and Effects Analysis (FMEA) is an engineering technique for identifying what can go wrong in a system and what happens downstream. Applied to emergency preparedness with AI, it transforms vague "what if" thinking into systematic vulnerability identification. Instead of hoping your emergency plan works, you stress-test it: ask AI to assume each component fails and trace cascading problems.

The traditional FMEA process assigns severity ratings (how bad if it fails?), occurrence likelihood (how probable?), and detection difficulty (would you catch it before disaster?). When applied to family emergency plans, this becomes concrete. Your emergency plan depends on your phone having battery. Severity: critical (no way to receive updates). Occurrence: moderate (phones die in extended outages). Detection: high (you'd know immediately). Result: you should add hand-crank radio and pre-established meeting points that don't depend on phone contact.

AI accelerates FMEA by systematically exploring failure scenarios. Rather than brainstorming what might go wrong, you structure a prompt: "Here's my family emergency plan: [paste plan]. For each component—communication chain, evacuation route, emergency contacts, resource stockpile—assume it fails completely. What downstream problems emerge? Which failures cascade? Which are recoverable?" The model generates a failure matrix. This is more thorough than human brainstorming because AI doesn't skip low-probability scenarios or assume everything will work out.

Real example: You have an emergency contact tree where one person is responsible for calling multiple family members. FMEA analysis reveals: if that person's phone dies, or they're injured, or they're inaccessible, nobody gets notified. This isn't a minor issue—it's a critical single point of failure. The AI flags it, and you redesign the tree with redundant contact paths. This is where systematic analysis beats informal planning.

Quantitative vs. qualitative FMEA differs in how you weight risks. Quantitative asks: probability of failure × severity of consequence = risk score. A low-probability, high-severity failure (house burns down, evacuation plan inaccessible) might score higher than high-probability, low-severity failure (phone battery dies, you use backup communication). Qualitative assessment is simpler: rank failures as critical, major, or minor. For personal safety, you probably want a hybrid—some risks are obviously critical (inability to contact emergency services), while others depend on your specific situation.

Integration with emergency contact chains is particularly valuable. Your contact chain only works if every link holds. AI can model this: "Here's my family contact tree. Person A calls B and C (2-minute window). B calls D and E. C calls F. If B doesn't answer, does D get notified? If the call chain exceeds the window (say, everyone is slow to answer), does someone get left out?" Forces you to add redundancy: parallel chains, predetermined time limits, escalation if contact fails.

Implementation with tools like ChatGPT or Claude involves structured prompts. Provide your complete emergency plan, then ask specific failure scenarios: "Assume I cannot access my home for 48 hours after evacuation. What items should I have in my 'go bag' that you don't have listed? What documents should I have copies of that aren't listed?" The model generates gaps by reasoning through consequences. You iterate: "Are these realistic gaps, or are you over-preparing?" Refine together.

The key insight is that FMEA isn't about catastrophizing—it's about pragmatic hardening. Some identified vulnerabilities will be acceptable (low probability, acceptable consequence). Others demand mitigation. By seeing the full failure mode matrix before crisis, you make conscious trade-offs instead of discovering gaps during actual emergencies.

Try this: Write a one-paragraph summary of your current emergency plan. Paste it into Claude with this prompt: "Assume each part of this plan fails. For each failure, describe what happens next and what problems cascade. Prioritize by how catastrophic the cascade is." You'll get a structured list of vulnerabilities ranked by impact. Take the top 3—those are your highest-priority improvements.