How can you prompt AI to draft a safety-incident report summary

The Prompt

Summarize this safety incident into a formal report.

Raw account: [what happened, in the order it occurred, from witness/involved-party notes]
Cover: what happened (factual, no speculation about cause), immediate response taken, and any immediate corrective action already implemented.
Keep root-cause analysis and disciplinary conclusions out of this summary — those come from the formal investigation process, not this report.

Why This Prompt Works

Incident reports that speculate about cause or assign responsibility before a formal investigation has actually happened create real problems — factually, they can be wrong, and procedurally, most safety programs require root-cause analysis to follow a defined investigation process rather than being baked into the initial incident summary.

How to Customize It

Have the actual witnesses/involved parties supply their account directly rather than letting the model infer sequence or cause from incomplete notes — accuracy of the factual record matters most in a document like this. Note the exact time and location precisely, since those details are often the ones a later formal investigation needs and are hardest to reconstruct after the fact.

Common Mistakes to Avoid

Don’t let the summary drift into “probably caused by” language — that’s exactly the speculation the instruction is designed to keep out, and it can complicate the actual investigation and any related liability process. Also avoid delaying the write-up until memory has faded; a same-day factual account is far more reliable than one reconstructed days later.

Bottom Line

A safety-incident summary should report only what’s known factually and what was done immediately after — cause and responsibility conclusions belong to the formal investigation, not this document.