Voice incident reporting for site workers in Australia
Short answer
Haznote lets a worker report a hazard by speaking. One control starts the recording and one stops it. The app flags a short or poor recording before it goes. Photos go into the same report, and the AI drafts the written record from the voice note for a manager to confirm.
Why voice
You can describe a hazard most clearly the moment you see it, and it’s hard to type with gloves on. Speaking takes seconds and keeps the detail. It also suits people who are quicker talking than writing. The recording is the worker’s own account, not someone’s summary of it.
What the worker does
- Tap the mic and describe what you see and where
- Tap stop when you’ve said enough
- Add photos in the same report
- Send, or let it wait if there’s no signal
Before the report goes, the app checks the recording. A dead microphone or clipped audio gets flagged. A very short take gets a question: record again, or save anyway? Those are the usual reasons a report turns out useless a week later, when nobody can ask the worker what they meant.
What the manager gets
The AI transcribes the note and drafts a formal description, a WHS category and a suggested priority. The manager sees the draft beside the worker’s original words and confirms or edits it. The AI’s draft is stored separately from the manager’s edits, so the audit trail shows both.
Limits, stated plainly
Voice recognition can mishear names, places and technical terms. That’s why the original recording and the original words stay attached to the draft, and a manager can listen to the recording. Haznote doesn’t claim to understand every accent or every noisy site. The manager decides what the report means.
How it works.
- 01 / Record
Mic, speak, stop
One control starts the take and one ends it. The worker describes what they see.
- Mic
- Stop
- 02 / Check
Problems are flagged
A dead microphone, clipped audio or a very short take is flagged before sending.
- Quality check
- 03 / Draft
AI drafts, a manager confirms
The written draft sits beside the worker’s own words until a manager signs it off.
- Transcript
- Draft
Plain answers.
Does it work in a noisy site?
The app flags a dead microphone or clipped audio, but it can’t promise perfect recognition on a loud site. The original recording stays with the report so a manager can listen.
Can a worker type instead?
Yes. A written report is supported alongside voice.
Who decides the classification?
A manager. The AI suggests, and a manager confirms or edits it.
Where this comes from.
This page describes how Haznote works. It makes no claims about the law, so it cites no regulator.
Try it on one real site.
Run it on one site and see the first report land in the manager’s list.