AI photo analysis and executive summaries

On-site photo analysis and executive-summary drafting are offered for every one of the 13 hazard types, each grounded in that hazard’s own methodology rather than a one-size-fits-all prompt.

What it is

Each hazard type has its own AI preset, and each preset carries a fixed, closed list of result values its photo-analysis feature can choose from — never free-text speculation, and never a value outside that list. Every enum restates only what’s visually present in the frame — an alarm, a detector, a crack, a socket, a reading having been photographed — and none of them names a material, a code, a measurement or a determination. Asbestos’s own result values describe only the visible condition of an already-identified or presumed material (damage, visibility), never an identification; radon’s describe only the detector’s visible condition and placement, never a reading or a concentration; indoor air quality’s describe only what instrument or feature is visible, never a verdict on air quality itself. Executive-summary drafting works the same way across every hazard type: it’s grounded in the specific readings and findings already recorded for that hazard, and it drafts a starting-point paragraph for a human to edit, never a paragraph that invents a fact not already on file.

How to do it

1. On a room or finding photo, use the on-site Analyze (or equivalent) button where your organisation has connected an AI provider — this returns a suggested value from that hazard’s own fixed list, for you to accept, adjust or dismiss.
2. Use the executive summary draft option from a survey’s own findings once enough has been recorded — it reads only what’s already on file.
3. Review every suggestion against what you can actually see or measure before accepting it — nothing is written into the sealed record until you do.
4. If a suggestion looks wrong for the hazard type, check the job’s hazard type is set correctly first — the wrong hazard type means the wrong result vocabulary is in play.

How it integrates

Photo analysis is only offered on a hazard type once that hazard declares its own result-value list — every one of the 13 shipped hazard types now does, including excess cold, which only gained its own preset and result contract very recently. The result vocabulary a preset returns is byte-identical to the enum embedded in that preset’s own prompt, so what the feature can say and what it’s instructed to say are kept in lockstep by construction, not by convention. Neither feature ever performs a statutory determination, identifies a material, reports a measurement or assigns a code — see How AI is used, and its boundaries for the drafts-only rule this all sits inside, and Customising AI prompts if your organisation wants to adjust the instruction text itself (the result vocabulary stays fixed either way).

Common problems

• A hazard type I expected to have no AI support now does. AI photo analysis was extended to every shipped hazard type, including ones that previously shipped without it — check the on-site button rather than assuming a hazard type has none.
• The suggested result doesn’t match what I can see. Treat it as a starting point, not a finding — adjust or dismiss it; nothing is saved until you accept a value.
• No analysis button appears at all. Your organisation hasn’t connected an AI provider — see Connecting your AI provider. The rest of the wizard works identically without one.