45+ AI-Drafted Publications for a National Safety Regulator

Our client is a national health and safety regulator responsible for helping businesses understand their obligations and improve safety in the workplace. Guidance is the primary way a regulator of this kind turns legislation into something an employer can actually act on, and the organisation produces more than 100 guidance documents each year to do it.

Developing that guidance is specialist work. Subject matter experts must interpret legislation, standards and industry best practice before they can produce a first draft, and for complex publications that process can take up to six months. The bottleneck is the scarcest resource the regulator has: expert time, spent assembling and structuring source material rather than applying judgement to it.

Maintenance is the second half of the problem. The organisation also maintains a library of more than 400 guidance documents, and legislation, standards and industry practice all keep moving underneath it. Keeping that content current is resource-intensive, and it is difficult to consistently identify which guidance most urgently requires review, which means a document can drift out of date without anyone noticing.

DataSing partnered with the regulator to develop a suite of AI assistants using Azure AI Foundry and Azure OpenAI, supporting both sides of the problem: the creation of new guidance and the ongoing maintenance of what is already published.

The AI drafting assistant generates structured first drafts from user prompts. It automatically references the relevant legislation and standards, and formats the output to the organisation’s own templates, so what an expert receives is a properly structured starting point rather than a blank page.

Alongside it, an AI review capability continuously compares published guidance against updated legislation, revised standards and emerging industry practice. It identifies outdated content, highlights inconsistencies between documents, and prioritises the library so the team can see which guidance needs attention first, instead of relying on periodic manual sweeps.

The solution was delivered iteratively. It initially supported simple guidance documents, then expanded to generate complex multi-section publications, incorporate stakeholder feedback, include technical diagrams and legal references, and identify emerging industry trends before they are reflected in formal standards.

The platform has generated more than 45 first drafts, and is expected to reduce initial drafting effort by a factor of four. Work that consumed months of specialist time now begins from a structured, referenced draft.

The review capability has identified dozens of high-priority guidance updates that would likely have remained undetected through manual review alone, turning currency from a periodic project into a continuous signal.

By combining AI with expert oversight, the regulator can produce guidance faster and maintain higher-quality publications, while focusing specialist expertise on the areas where it delivers the greatest value. Experts still make the judgement calls; the AI removes the assembly work in front of them.