The problem with traditional risk assessment
Until recently, producing a risk assessment meant one of three things: hiring a health and safety consultant, spending hours on a blank form, or downloading a generic template and filling in the same boilerplate every business in your sector uses. All three have the same weakness — the output is only as good as the knowledge and effort put in, and for most small businesses, both are in short supply.
The HSE’s own research consistently shows that smaller employers struggle most with risk assessment quality. The most common failures are generic language that could apply to any business, failure to identify hazards specific to the actual activity, and control measures listed without any explanation of how they will be implemented. These are not just quality problems — they are legal ones. A risk assessment that is not ‘suitable and sufficient’ under Regulation 3 of the Management of Health and Safety at Work Regulations 1999 is no risk assessment at all.
What AI actually does
AI tools like Anyrisks work differently from a template. You describe your specific activity — the task, the location, the sector, the people involved, any hazards you are already aware of — and the tool produces a written assessment tailored to those facts. It identifies the hazards likely in that type of work, names the specific UK regulations that apply, applies the hierarchy of controls in the legally required order, and produces a document you can review, sign off, and put into use.
The practical effect is significant. A risk assessment that previously took hours to produce — researching applicable regulations, drafting hazard descriptions, identifying who is at risk, working through control measures — can be produced in under two minutes. The quality ceiling is also higher: the AI draws on knowledge of sector-specific regulations and HSE guidance that most non-specialists would not have at hand.
What has not changed
The legal framework has not changed. The duty to carry out a ‘suitable and sufficient’ risk assessment under MHSWR 1999 Regulation 3 still sits with the employer or self-employed person. That duty cannot be delegated to a software tool. An AI-generated assessment must be reviewed by a competent person before use — someone who can confirm it reflects the actual workplace, that no significant hazards have been missed, and that the control measures described can and will be implemented.
This review step is not a formality. AI tools work from the information provided. If you describe a building site but do not mention that work will take place near overhead power lines, the output will not cover that hazard. The competent person reviewing the document is the check that catches gaps between what was described and what is actually true on the ground.
Where AI adds the most value
The biggest gains are for businesses that produce risk assessments frequently but do not have in-house health and safety expertise. Sole traders, small contractors, nurseries, event organisers, market traders — anyone who needs a credible, regulation-referenced document for a specific job or activity, without paying consultant rates or spending half a day on paperwork.
AI also helps with consistency. When a business produces risk assessments across multiple jobs or sites, manual processes produce variable quality. An AI-generated process produces documents at a consistent level, covering the same regulatory ground each time, with a format that is legible and complete.
How to use an AI-generated risk assessment correctly
The single most important factor is the quality of the information you provide. The more specific your description of the activity and the location, the more accurate the output. Vague inputs produce generic outputs — exactly the problem AI is supposed to solve. Describe the actual task, name the actual location, mention any specific hazards you already know about.
Once generated, review the document before use. Check that the hazards listed reflect your actual workplace. Check that the control measures are ones you can genuinely implement. Add anything the AI has missed. Then sign it, keep it on file, and review it when circumstances change — exactly as you would with any other risk assessment.
Further reading
For the legal requirements in full, see our risk assessment legal requirements guide. For the complete 5-step process and hierarchy of controls, see the ultimate guide to risk assessment. For a practical walkthrough of using Anyrisks, see how to use AI for instant risk assessment.
