AI is coming to occupational health whether we are ready or not. In fact, it is here already in some of our workflows and clinical tasks. I work in the NHS, and the NHS 10 Year Plan commits £10bn to digital transformation and names AI as one of five transformative technologies. But what happens when the algorithm works and the system around it doesn’t? This may sound like a familiar problem in healthcare.

That question was the focus of a recent ACPOHE webinar I delivered, and it is the question I want to explore here.

Where are we now?

We recently surveyed UK occupational health professionals about their use of AI. The results tell a clear story. Sixty per cent of respondents have experimented with AI tools, mainly for documentation and research. But only twelve per cent use AI in clinical decision-making. The technology is here, but it is not yet transformative.

What is holding us back? Not the technology. The biggest barriers are human: data security (72% concern), regulatory uncertainty (68%), and professional autonomy (61%). We are not Luddites in OH, we are future and work focussed. Ninety per cent of us want AI-specific training, and 75% want clearer regulatory oversight. We see the potential, but we also see the risks.

A simple framework for evaluating AI

Jessica Morley, a researcher at Yale Digital Ethics Center, has developed a simple framework for understanding the risks of AI in healthcare. I have found it useful for thinking about AI in OH.

• Epistemic risk: AI gives wrong or biased answers. For example, a return-to-work prediction tool trained on large corporate data will fail for SME or gig economy workers.
• Normative risk: AI changes practice for the worse. An AI that writes consultation notes might look efficient, but if clinicians stop taking detailed histories because “the AI will capture it,” we lose something essential.
• Traceability risk: No one is accountable when AI fails. If an AI-assisted recommendation leads to harm, who is responsible? The employer? The OH professional? The vendor?

These are not technical problems but governance issues.

Practical tools for everyday AI use

If you are using AI for drafting, summarising, or research and many of you are, here are two frameworks to help.

The 4 Cs of effective prompts:

• Clear Role & Context: “You are an OH advisor with ten years of NHS experience…”
• Specific Constraints: “Use only NICE and HSE guidance from 2020 onwards…”
• Clear Output Format: “Provide a bullet-point summary with key findings, clinical implications, and limitations…”
• Critical Evaluation: “What are the limitations of this advice? What should I verify before using it clinically?”

The CARE Framework for evaluating AI outputs:

• Check the source – are references real?
• Assess clinical reasoning – does it align with your knowledge?
• Review for context – is this relevant to your setting and population?
• Evaluate currency – is the guidance up to date?

What you can do next week

You do not need a Trust-wide or private employer led strategy to start. For yourself: try using a free AI tool to draft one report or summarise one research paper. Then verify everything. See where it helps and where it fails. Document your use in your CPD portfolio.

For your team or service: ask whether your organisation has a policy on staff use of AI tools. Ask whether any AI tools have been procured without a clinical safety or data protection assessment.

For independent practitioners: review your professional indemnity insurance – does it cover AI-assisted decisions? Review your data protection registration – does using a cloud-based AI tool require a Data Protection Impact Assessment?

The bottom line

AI is here to stay. Whether it helps or harms depends on the governance we build around it. The technology is not the hard part. The system around it is and our job as OH professionals is not to learn to code. It is to ask the right questions about data, about accountability, about bias and to build the governance that makes AI safe for the workers we serve. Within iOH, we have started work on this and the collaborative group is making small waves in the AI in OH ocean.