AI workplace hazard analysis

Every photo of the floor is a hazard report waiting to be read.

Upload one image. Get back ranked hazards, the root cause behind them, and the controls that fix them.

We don't store your photos on our servers. Images are processed for the analysis request and kept only in your browser unless you export or email a report. How image privacy works →
Who this is for

Built for EHS leads, production managers, shift leaders, maintenance teams, and trainers who need faster hazard spotting, better toolbox talks, and cleaner corrective-action reports.

EHS leads Production managers Shift leaders Maintenance teams Trainers
One pass, four outputs

It doesn't just spot hazards. It explains them and tells you what to do.

The same flow a seasoned EHS lead runs in their head — identification, root cause, controls — done in seconds and written in standards language you can paste into a report.

01 / IDENTIFY

Ranked hazards

Each hazard scored by severity and likelihood across the PESTLE-H taxonomy, with the people at risk and the injury mechanism named.

02 / TRACE

5 Whys root cause

Drills past "didn't follow procedure" to the management-system or design decision that made the hazard inevitable.

03 / CONTROL

Hierarchy of Controls

Actions ordered Elimination → PPE, each with an owner, a timeline, and the primary standard (OSHA / EN / ISO) cited.

04 / BRIEF

Export & share

A Swiss-Cheese note on defeated defences, a supervisor takeaway for the next toolbox talk, and export to a PDF report, One Point Lesson, or email (see below).

Take it with you

Every analysis, in the format your site actually uses.

The result isn't trapped on screen. Turn any analysis into a document you can file, print, pin to the line, or send — in one click.

PDF REPORT

Full hazard report

A polished, multi-section PDF: the annotated photo with numbered hazard markers, every ranked hazard, the 5 Whys, a "What Could Happen" summary, and a controls table with owners, timelines, and the cited standard. Ready to attach to an incident record, audit, or risk assessment.

OPL

One Point Lesson

A single-page, visual-first shop-floor sheet — the photo, the top hazards, and the key action, sized to print and pin by the line for a toolbox talk. The OPL number is left blank so you can log it in your own numbering system.

EMAIL

Send it on

A formatted summary that opens straight in your mail app, ready to send to a supervisor or EHS lead — no copy-pasting, no reformatting.

COPY / TEXT

Into your own system

One-click copy or a plain-text download to drop the findings into your CMMS, spreadsheet, permit, or ticketing system.

Every export is built from the interactive hazard map — numbered markers you can nudge to the exact spot, with the photo shown full-screen when you need the detail. Analysing several photos at once? They come back as one combined PDF.

AI makes mistakes, and this tool is no exception — it does the majority of the work and is a great way to learn, but every report is a strong first draft, not the final word. Please review before sharing.

Built on the canon, not vibes

Blame-free by design. Systems-first by default.

Every analysis runs through established safety-science frameworks and resolves to systemic causes, never individual fault — so the output holds up in a real investigation.

PESTLE-H taxonomy 5 Whys Hierarchy of Controls Swiss Cheese model Just Culture OSHA · EN · ISO standards
🔒

How image privacy works

We don't store your photos on our servers. Here's exactly what happens to an image, step by step.

Stays on your device
  • Loaded into browser memory only — never written to disk by this app
  • No image database — nothing to breach, leak, or subpoena
  • Reports stay in your browser until you export or email them
Leaves your device
  • Sent to Anthropic's API over HTTPS to perform the analysis
  • A resized copy is sent separately if you use Auto Blur
  • If you email a report via Resend, the image is embedded in that email

Under standard Anthropic API terms, inputs aren't used to train models or retained beyond the request — verify against your own account terms if you're uploading sensitive site photos. Prefer everything client-side? Bring your own Anthropic key and calls go straight from your browser to the model.

Why I built it

Built for my worst days on site.

I spent years watching the same thing play out: a good supervisor could glance at a line and clock hazards a new hire would walk straight past — but that instinct never left the building, and it usually surfaced too late, as blame, after someone was already hurt. I didn't want another clipboard or a course nobody remembers. I wanted that expert eye in everyone's pocket. So I built SafetyPulse — image analysis wired to the frameworks I trust (PESTLE-H, 5 Whys, Hierarchy of Controls), with one line I've never been willing to cross: find the system that failed, never the person.

And I'll say the quiet part out loud: AI must never take away accountability, responsibility, or personal connection — the walk-and-talk, the judgement call, and the person who signs their name to the outcome. SafetyPulse is built to support those, not replace them. It reads the scene and names the hazard; you confirm it, own it, and act on it. The tool sees — a person is still responsible.

And it won't be perfect — AI makes mistakes, and this tool is no exception. But it does the majority of the work, and it's a genuinely good way to learn to spot hazards and understand why they matter. Treat it as a sharp second pair of eyes, not the final word.

Pay for what you analyse

Start free. Buy credits only when you need them.

No subscription. A flat $0.79 per analysis, and credits never expire. Bring your own Anthropic key and analyses are half price.

Start here

Free

$0
Start with 3 free hosted analyses — no card required.
Or bring your own Anthropic API key for lower-cost private analysis.
  • Every framework, every export
  • Auto-blur faces & text
  • No credit card to begin
Create account

Standard

$23.70 / one-time
30 analyses at $0.79 each. For a safety lead running regular walkthroughs.
  • Credits never expire
  • Pay-as-you-go, top up anytime
  • Full export & email reports
Get Standard
Bring your own key

BYOK

$0.40 / analysis
Half price. You supply the Anthropic key and pay Anthropic directly for the AI; we charge a reduced platform fee.
  • Your key & image stay in your browser
  • Half the standard per-analysis price
  • Same credits, twice as far
Use my own key

Also available: Starter $7.90 (10 analyses) · Pro $47.40 (60 analyses, up to 10 images per batch) — choose at top-up.

Next shift starts now

Find the hazard before it finds someone.

Point your phone at the work, upload the photo, and read what the scene is telling you.