How to Use ChatGPT to Draft SWMS
In Australia, high-risk construction work can't start without a Safe Work Method Statement (SWMS). Too often, the one on site is last job's SWMS with a new project name. Tell ChatGPT about the work, the site and your crew, and you'll have a draft written for this job in minutes. Check it against the site, then have a competent person sign it off before work begins.
Trusted by companies, projects and teams of all sizes
The short answer
Give ChatGPT the details of the high-risk work, the site and the crew, and it'll draft a SWMS right in the chat. You'll get job steps in order, hazards for each step, risk ratings before and after controls, the controls themselves and a sign-on section. A competent person then checks it against the site, makes any fixes and signs it off.
How to Get ChatGPT Drafting Your SWMS
You don't need a long prompt. Give ChatGPT the specifics of the work and the sections a SWMS needs, then tighten the draft it returns. After that, the same prompt gets reused and only the work and site change.
-
Step 1
Give ChatGPT the Work and the Site
ChatGPT's draft is only as specific as what you give it. Name the task, the site, the plant and the crew, then say what makes this job different. Roof sheeting over a fragile skylight, a trench beside a public footpath or an EWP working near power lines will each get you a far more useful SWMS than "a roof job".
- The high-risk construction work, named plainly
- The site and its conditions on the day
- The plant, the access and any permits needed
- Who's exposed: the crew, other trades and the public
-
Step 2
Tell It the Sections a SWMS Needs
Left alone, ChatGPT writes prose. Give it the sections by name and it follows them, so the draft comes back in the format your principal contractor accepts.
- Job steps in the order the crew works
- Hazards for each step, specific to the work
- A before and after risk rating on your own matrix
- Controls in hierarchy order, with PPE last
- A sign-on section for every worker
-
Step 3
Tighten the SWMS to the Site
Treat the first draft as a starting point. Cut anything generic and add the hazard that's actually on site today. Check the steps match how the crew will really do the job, then the supervisor or safety officer reviews it and signs it off.
Prompt Guidance for Drafting SWMS with ChatGPT
Build a good SWMS prompt once and reuse it, changing only the work and the site. Name the task, the plant and the site every time. The more specific you are about the work, the less generic the SWMS that comes back from ChatGPT.
- "Draft a SWMS for [high-risk work] at [site] using [plant]. Set it out as job steps in order, with the hazards for each step, a risk rating before and after controls, and the controls."
- "We're replacing roof sheeting from an EWP on a warehouse roof. There's a fragile skylight, and power lines run along the rear boundary. Draft a SWMS for this job."
- "Rate each hazard on a 5x5 likelihood and consequence matrix, then show the residual risk after controls."
- "List the controls by the hierarchy of control, with PPE as the last resort. End with a sign-on table (name, company, date, signature) and a supervisor approval line."
- "Use the same structure for a new SWMS: [task] at [site]. Keep the hazards specific to this job and list any assumptions you've made."
Troubleshooting Prompts for Drafting SWMS
A weak draft usually comes from a thin prompt or an under-structured template, and most of the fixes sit in the template. Work through them in the order below.
- The SWMS reads as generic: the site is missing. Add the location, the plant, who's working nearby and where the public is.
- The hazards are wrong: ChatGPT had nothing to draft from. Paste in your existing SWMS or the relevant safe work procedure section.
- Risk ratings don't match yours: it's using its own matrix. Write your risk matrix into the template and paste it into the prompt.
- Sections go unused: ChatGPT is guessing the structure. Give it your template headings by name.
- There's no sign-on section: you didn't ask for one. Ask for a table with name, date and signature columns.
- Controls drift away from the steps: the layout allows it. Ask for one row per step, with its hazards, rating and controls.
Limitations
ChatGPT drafts the SWMS and can help polish the wording, but it doesn't run the job. A competent person reviews and signs it off before work starts, and remains accountable for every control.
What ChatGPT Does Well When Drafting a SWMS
The more detail you give it, the better the draft.
- Drafts a complete SWMS from your description of the work
- Lists job steps in order with hazards and controls for each
- Suggests risk ratings and controls you can accept or change
- Follows the headings in your own SWMS template
- Rewrites sections in plain language for the crew
- Tightens the wording for the principal contractor
What ChatGPT Leaves to You When Drafting a SWMS
A ChatGPT draft is a starting point, not a finished SWMS. The rest is up to your team.
- Knowing the site, since it only works from what you tell it
- Judging whether controls are adequate, which is the competent person's call
- Signing off the SWMS
- Guaranteeing WHS compliance
- Collecting crew sign-on and running the job
What This Assumes You Already Do
This page assumes your team already writes a SWMS for high-risk work and gets it signed. ChatGPT only drafts the document. You still need the habit of reviewing it, and a way to capture sign-on.
- Your team already writes a SWMS before any high-risk construction work
- Your SWMS format is one your principal contractors accept
- A competent person reviews and signs off each SWMS
- Every worker signs on before starting, and signs again when the SWMS changes
Why Use ChatGPT to Draft SWMS
Most crews know how to do the job safely but don't have time to write it all down. The job starts Monday, the principal contractor wants a SWMS before anyone gets through the gate, and the quickest option is last job's version with a new name. That leaves the crew signing a document written for different work. ChatGPT gives you a tailored starting point in minutes, so your competent person can spend their time on the hazards that are actually on site.
What a Good SWMS Contains
A SWMS that meets this bar is one your principal contractor will accept and your crew will understand. Check every point below before the competent person signs it off.
- The high-risk construction work, named plainly
- Job steps in the order the crew actually does them
- The hazards for each step
- A risk rating before and after controls
- Controls that follow the hierarchy of control, with PPE as the last resort
- How each control gets checked on the day, and by whom
- The preparer and approver by name, plus every worker who signed on
The principal contractor wants the SWMS before the crew's allowed through the gate at six. The only one on the laptop was written for a different job.
Connecting Sitemate MCP to ChatGPT
Sitemate MCP lets ChatGPT read the signed SWMS you hold in Dashpivot. It's read-only, so ChatGPT can't create, edit or delete anything. Connecting takes three steps, and the first happens outside ChatGPT: an Org Controller turns on MCP access in Dashpivot. You'll also need ChatGPT Plus, Team or Enterprise, because connectors aren't available on the free plan.
-
Step 1
Turn On MCP Access in Dashpivot
An Org Controller goes to Dashpivot Settings > Security, sets MCP Access to anything except None and saves. The app won't authenticate until this is done.
-
Step 2
Add the Sitemate App in ChatGPT
In ChatGPT, go to Settings > Apps and Connections > Explore Apps. Search for Sitemate, open the result and click Connect. Sign in with your Sitemate credentials and approve the permissions.
-
Step 3
Check the Tools Are There
Open a new chat in ChatGPT. The Sitemate tools appear under the tools icon. They can read your records, but they can't create, edit or delete anything.
If the App Won't Appear
Three causes account for almost every failed connection, and the first is by far the most common.
- Nothing connects: MCP Access is still off in Dashpivot, and only an Org Controller can turn it on
- Sitemate isn't in the app list: connectors need ChatGPT Plus, Team or Enterprise, not the free plan
- The tools are missing in chat: start a new chat and check the tools icon
How the SWMS Gets Captured
ChatGPT's part ends once the draft is written in the chat. From there Storm, Sitemate's data capture agent, imports it into Dashpivot and captures it on site, and Sitemate MCP reads the signed SWMS back.
-
Storm Imports the Draft into Dashpivot
Save the ChatGPT draft as a PDF, Word or Excel file, and Storm rebuilds it as a Dashpivot SWMS template with AI Templates. The supervisor or safety officer checks the template before it reaches the crew.
-
Storm Captures It on Site
The crew opens the SWMS on a phone or tablet, and Storm fills in the site detail with AI Form Fill, from what they say and photograph. The supervisor checks it and every worker signs on before starting. The signed SWMS is then held against the job in Dashpivot.
-
Sitemate MCP Reads It Back
Once your SWMS are captured in Dashpivot, ChatGPT can read across them through Sitemate MCP. Ask which jobs have a signed SWMS, or which workers haven't signed on yet. It's read-only and your user permissions still apply, so ChatGPT only sees what you can. More on that in How to use ChatGPT to review SWMS.
Storm Captures the SWMS ChatGPT Drafts
The person doing the work describes the job and takes photos as they go, and Storm fills the SWMS from that. The supervisor checks it and the crew signs on before starting, so nothing waits for the end of the day.
AI Form Fill
Capture what you see and say, and Storm drafts the form for you. The details land in the right fields, so there is a record to check the moment the work is done, and you review and submit it yourself.
Listens to the work
Captures voice cleanly through machinery, wind, and movement. Accents and trade language understood. No headset needed.
Works on site
Built for gloves, cabs, and screen protectors. Captures inputs offline and processes in the background. No waiting on the field.
You stay in control
Storm never commits a record on its own. Every output is a draft for human review. Admins control what Storm can write to.
Frequently Asked Questions
Can ChatGPT write a SWMS for me?
It can write the first draft. Describe the high-risk work, the site, the plant and the crew, and it’ll come back with job steps, hazards, risk ratings and controls. A competent person still reviews and signs it before anyone starts.
Does ChatGPT know the WHS rules in my state or territory?
Mention your state or territory in the prompt and ChatGPT will draft with those rules in mind. Treat it as a starting point. A competent person still needs to check it against the WHS duties that apply.
Can I match the draft to our own SWMS format?
Yes. Share your headings and risk matrix, and ChatGPT will write to them, sign-on table included, so the draft fits the template your principal contractors already accept.
What does ChatGPT see of our SWMS and site data?
While drafting, it only sees what you type or paste in. If you connect Sitemate MCP, it can also read the SWMS records your Dashpivot permissions allow. The connection is read-only, so nothing can be created, edited or deleted. You’ll need ChatGPT Plus, Team or Enterprise to connect.
Where does the SWMS get signed on and stored?
On site, in Dashpivot by Sitemate. Save your ChatGPT draft as a PDF, Word or Excel file, and Storm, Sitemate’s data capture agent, rebuilds it as a template with AI Templates. Your crew fills it in and signs on in Dashpivot, where it’s stored against the job.