thirdbrain.

What's possible with AI in WHS and safety management

Evidence current to mid-2026.

In Work Health and Safety (WHS), AI is ready today for the writing work: toolbox talks, induction scripts, incident report first drafts, and safety performance summaries, always with the safety lead reviewing and making the safety calls. Risk rating, approving controls, determining notifiability, and return to work (RTW) medical decisions stay with people. The tables below show exactly which is which, with Australian law as the line.

Each task below sits in one of three bands: a strong fit today, worth a careful pilot, or keep with people for now.

A clear breakdown of where AI does and does not fit across the core tasks of WHS and safety management. Based on Australian law and recent evidence, not vendor marketing. Throughout, ‘the safety lead’ means whoever holds the WHS duty at your business: a dedicated WHS officer or safety manager, a site manager with WHS responsibilities, or the business owner.


Where AI is mature: strong fit today

Proven and available today. AI does the bulk of the work and the safety lead reviews it.

FunctionThe job todayWith AI
Toolbox talk and induction contentWrite site-specific toolbox talk scripts and induction briefings, translate for diverse workforcesAI drafts topic-specific scripts and multilingual versions; the safety lead checks site accuracy and Australian regulatory alignment before anyone hears it
Incident report first draftTurn field notes, photos, and verbal accounts into a structured written reportAI drafts from notes, photos, or audio; the safety lead verifies the facts, determines whether the incident is notifiable, and signs off
Training records and competency trackingMaintain a training matrix, track certification expiries, ensure every worker holds the required ticketsSoftware tracks expiries and flags gaps 30-90 days out; the safety lead follows up and verifies licences against regulator records
Safety performance reportingCompile incident counts, Lost Time Injury Frequency Rate (LTIFR), near-miss rates, and write the management safety summaryAI assembles the data and drafts the narrative; the safety lead verifies every number, owns the interpretation, and does not let AI assert what caused the trend

Where AI is emerging: consider piloting with a human gate

Promising but not yet proven at this scale. AI assists and the safety lead stays in the loop, so trial it on a contained scope first.

FunctionThe job todayWith AI
SWMS draftingWrite a Safe Work Method Statement for each type of high-risk construction work, listing every hazard and controlAI generates a first-draft SWMS from a description of the work; a competent person reviews (AI tends to underweight engineering controls), worker consultation happens, and the PCBU owns the final document
Risk register maintenanceIdentify workplace hazards, list them in the register, assess likelihood and consequence, assign and review controlsAI prompts hazard identification and helps structure the register; a competent person rates every risk and approves every control
SDS and chemical register managementMaintain a register of every hazardous chemical with a current Safety Data Sheet (SDS) (no older than five years), flag incompatible storageAI tools extract data from SDS documents and flag gaps; the safety lead verifies critical fields, does the physical site walkthrough, and checks Globally Harmonised System version 7 (GHS 7) compliance
Inspection and audit checklistsBuild site-specific inspection checklists, run the inspection, track corrective actions to closeAI generates checklist structures and can summarise recurring findings across audit sets; the safety lead does the physical inspection and verifies corrective actions in the field
Safety management plan draftingWrite or update the WHS management plan covering responsibilities, hazard management, and emergency proceduresAI drafts the structure and standard sections; a competent person checks that higher-order controls are not being skipped and adds the site-specific content
RTW and workers comp adminLodge and manage the workers compensation claim, coordinate RTW plan documentation, draft suitable duties proposals, track recovery and ongoing treatmentAI drafts claim and RTW documents, tracks milestones and treatment, and reconciles treatment invoices against the plan; the safety lead owns the worker relationship and the insurer liaison, and only the treating doctor certifies work capacity
Contractor and visitor inductionBuild pre-qualification questionnaires, online induction modules, and site-specific visitor briefingsPlatforms manage pre-qualification status and track completion; AI drafts the induction content; the safety lead approves each contractor and supervises them in the field
Regulatory correspondence (routine)Draft routine enquiries and standard letters to WHS regulatorsAI drafts routine enquiry and acknowledgement letters; the safety lead verifies factual accuracy and signs. Formal compliance submissions stay with the PCBU or another responsible officer

Where AI is not ready or suitable today: keep with people

These stay with people, either because the tooling is not reliable enough or because the law and accountability rule it out.

  • Determining whether an incident is notifiable. Under WHS Act sections 35-38, the PCBU must assess whether an incident meets the statutory definition and notify the regulator immediately. AI can tell you what the criteria are; the safety lead makes the call and sends the notification.
  • Risk rating and approving controls. Likelihood and consequence ratings must be made by a competent person who knows the specific workplace, the workers, and the controls already in place. AI cannot inspect the site or consult workers as required by WHS Act section 47.
  • SWMS sign-off. The PCBU must ensure the SWMS accurately reflects the actual work, plant, and site. The PCBU or a competent person with knowledge of the specific job must review it. AI cannot.
  • RTW medical determinations. Work capacity certificates are issued by registered medical practitioners under the Australian Health Practitioner Regulation Agency (AHPRA). No AI tool can certify incapacity, determine suitable duties, or override a treating doctor’s certificate.
  • Workplace investigation findings. A credible investigation requires interviewing witnesses, assessing credibility, and weighing evidence under procedural fairness. AI can organise documents and draft timelines; the investigator reaches the finding.

Worth weighing: for any of these, the upfront setup, mapping site hazards, building templates, structuring the register, is a one-off cost. The ongoing job on the strong-fit items is mostly review, not the old grind. The time that comes back goes to the field: worker consultation, checking that controls are actually holding, and the judgement calls that only someone on the ground can make. AI takes the writing and the record-keeping; the safety lead stays where they are needed most.

This guide sits on top of the things that stay the same whatever your function. See the ground rules

Questions

The questions leaders ask.

Will AI replace our WHS officer or safety manager?

No, it changes what they spend time on. The writing, the record-keeping, and the document administration shrink, and their professional judgement, site knowledge, and legal accountability matter more. The calls that carry legal weight, risk ratings, notifiable incident decisions, Safe Work Method Statement (SWMS) sign-off, and investigation findings, stay with the safety lead. The goal is a safety function that does more of the right work, not one with fewer people.

Is it safe to use AI for risk assessments and SWMS?

It is safe as a drafting tool when a competent person owns the output. AI can generate a useful first draft of a SWMS or a risk register structure, but independent research shows AI tends to cluster recommendations at the procedural end of the hierarchy of controls, favouring training and signage over engineering interventions. A competent person must review the draft, consult workers as required by the WHS Act, and take responsibility for the final document. The person conducting the business or undertaking (PCBU) cannot transfer that obligation to software.

Where should we start?

Pick something narrow, repetitive and easy to check. Toolbox talk scripts are a common first step: AI drafts, the safety lead checks the site accuracy and Australian regulatory alignment, and the time saving is immediate. Incident report drafting from field notes is the other strong one, because turning scattered notes and photos into a clear written report is repetitive work, and the safety lead still verifies the facts and makes the notifiability call. Prove one before adding the next.

Can AI help us stay compliant with the WHS Act?

AI can help you document, track, and draft your way to better compliance. It cannot discharge a legal obligation that belongs to the duty holder. Under the WHS Act, the PCBU has a non-delegable primary duty of care, officers have personal due diligence obligations, and specific tasks such as risk assessments, SWMS, and notifiable incident decisions require a human with the right knowledge. AI is drafting and administration support; the PCBU and its officers remain accountable.

What is new in AI for WHS in the last six months?

Quite a bit, and it points one way: how you use AI is now part of your WHS duties. In NSW, the Work Health and Safety Amendment (Digital Work Systems) Act 2026 received assent on 18 February 2026, the first law in Australia to put explicit WHS duties around using AI, algorithms and automation to allocate or monitor work. The duties are not in force yet, they wait on a proclamation and on guidelines SafeWork NSW has been consulting on through the first half of 2026, so the time to write down your risk assessment is now. From 1 July 2026, NSW approved codes of practice, including the psychosocial hazards code, become benchmarks you must meet or better, and AI-driven monitoring and work allocation sit squarely inside those duties. Also from 1 July 2026, South Australia drops its high-risk construction height threshold from three metres to two, so an AI tool trained on the old rule will scope SWMS wrongly for SA sites. None of this stops you using AI for safety work. It means the safety lead stays in the loop, you record how you assessed the risk, and you treat what the tool produces as a draft to check.

How current is this, and what if the tools have moved on?

The evidence is current to mid-2026 and we refresh it as the tools and the rules change, which in WHS they do often. Where a finding comes from an independent source we lead with it; where it comes from a vendor we treat it as marketing. If something has shifted since you read this, the fastest way to get the current picture for your specific setup is a quick chat.

Book a discovery call

A first conversation

See how this applies in your world.

Book a confidential chat. Tell us where you see the opportunity, in whs and safety management or anywhere else, and we will agree a sensible next step. No deck, no pitch.