A General AI Tool Doesn't Know What Evidence to Collect. Your Investigation Does.
You described the incident. The AI tool organised your description into something that looked like a report. And somewhere in the back of your mind you thought: I wonder if I should have collected more.
That instinct is correct. Here's why.
This article is part of the series Why a Chat Window Isn't an Investigation.
The Tool Responded to What You Said. Not to What Happened.
A general AI tool generates a response based on your input. If you mentioned witness statements, it may have suggested you get witness statements. If you mentioned the floor was wet, it noted the floor was wet. What it cannot do is independently identify the categories of evidence that a complete workplace incident investigation requires — regardless of what you happened to include in your description.
This matters because the evidence you collect is the foundation of everything that follows. Your root cause analysis is only as strong as the evidence supporting it. Your report is only as credible as what's behind it. And the gaps in your evidence are exactly what insurers and regulators look for when they review an investigation.
If the evidence collection phase was guided by a chat prompt, those gaps are likely there. They're just not visible yet.
What a Complete Evidence Framework Looks Like
A structured workplace incident investigation uses a framework called PEEPO — People, Environment, Equipment, Procedures, Organisation — to ensure evidence is collected across every relevant category, not just the obvious ones.
Here's what that means in practice for a typical incident:
People covers witness statements, training records, competency assessments, and supervision records. Not just "who was there" but what each person's role was, what they were trained to do, and whether supervision was adequate.
Environment covers the physical conditions at the time — lighting, floor condition, layout, weather if relevant, any recent changes to the work area that may have contributed.
Equipment covers maintenance records, inspection logs, age and condition of any tools or machinery involved, and whether the equipment was fit for purpose.
Procedures covers whether a relevant safe work procedure existed, whether it was current, whether the worker had been trained on it, and whether it was being followed at the time.
Organisation covers scheduling, workload, staffing levels, and whether any organisational pressures may have contributed to the incident.
A general AI tool doesn't apply this framework. It responds to your description. If you didn't mention training records, it won't prompt you to collect them. If you didn't describe the maintenance history of the equipment involved, that gap stays invisible.
What this looks like in practice: A PEEPO-based checklist for a slip-and-fall incident surfaces evidence requirements across all five categories — including items the owner hadn't thought to mention in their original description of the incident, such as equipment maintenance logs, spill response procedures, and training records.
Want to see how a structured investigation handles this? Start your free trial
The Worked Example: What Was Actually Missing
A warehouse worker slips on spilled oil near a loading dock in a small manufacturing business and loses time from work. The owner describes the incident to a general AI tool and receives a response that captures the basic facts: a slip on a wet floor, the worker went to hospital, operations were disrupted.
The AI tool's response suggested witness statements from those who saw the incident and a note about posting a wet floor sign. What it never mentioned: whether the forklift operating in that area had a current maintenance record, whether the worker had completed manual handling and slip hazard training, whether there was a documented procedure for spill response in the loading dock area, and whether supervisors had recorded any prior observations about the floor condition in that zone.
None of those items appeared because none of them appeared in the owner's description. A purpose-built investigation tool generates a PEEPO-based checklist for the incident type and jurisdiction — flagging the evidence that matters for this specific scenario, regardless of what the owner thought to mention.
Why the Gap Matters Later
Evidence gaps don't create problems immediately. The report looks fine. The incident is filed. Then, weeks or months later, someone asks a question the evidence can't answer — an insurer querying whether training was current, a regulator asking about the maintenance log.
At that point, the evidence is often gone. The floor has been repaired. The equipment has been serviced. The witnesses have moved on. What wasn't collected at the time cannot be collected now.
A structured investigation collects evidence while it's available and flags what's missing before the investigation is closed. That's the difference between an investigation that holds up and one that falls apart when someone looks closely.
This evidence gap also flows directly into root cause analysis — One Prompt, One Answer. That's Not Root Cause Analysis. — because chains of causation can only be traced through the evidence that was actually collected.
If a regulator or insurer asks to see your investigation and key evidence was never identified or collected, that gap works against you. Not as a reflection of your intentions, but as a factual limitation of the record. Starting with a complete evidence framework is the simplest way to avoid that outcome.