Most compliance work has a well-worn path. A standard sets out the requirements, guidance explains them, and thousands of organisations have walked the route before you. The hard part is doing the work well.
Every so often, a team arrives with something that does not fit. A product that sits between categories. A model of care the standards never imagined. A new way of splitting responsibility between a supplier, a provider and a clinician. AI that changes how a decision gets made. The questions stop being “what does the standard say?” and become “what should the standard say about this?”
That is where we have done some of our most important work.
What makes a model novel
In our experience, novelty usually shows up in one of four places.
- Intended purpose. The product does something existing classifications were not written for, so whether it is a medical device, and what class, is genuinely open.
- Accountability. Clinical responsibility is shared in a new way, across organisations, roles or settings, so it is unclear who owns which risk.
- Change. The product learns or updates in ways a one-off approval was never designed to cover.
- Pathway. The technology moves care somewhere new, into the home, into a different profession, or across organisational boundaries, so the hazards are new too.
In each case, there is no template to copy. And the cost of getting it wrong is high in both directions. Too cautious, and a product that could help patients sits on the shelf. Too casual, and the first serious question from a regulator or a Clinical Safety Officer unravels months of work.
How we approach it
We have helped customers through exactly this kind of territory. Our recommendations were adopted, and they were the first instances of their kind that the NHS bodies and regulators involved had experience with. The response was overwhelmingly positive. The way we got there is worth sharing, because it is repeatable.
First, we start with deep understanding. Before anything else, we take time to understand the product, the people who will use it, the pathway it sits in, and the organisation behind it. Novel questions are rarely answered from the standard alone. They are answered from a clear picture of what the technology actually does.
Second, we reason from first principles. Every standard exists to manage a set of risks. When the letter of a requirement does not fit, the purpose behind it still does. We ask what the requirement is trying to protect, and design an approach that protects it just as well, or better.
Third, we make the rationale visible. A recommendation in new territory is only as strong as the reasoning behind it. We document why each decision was made, what alternatives were considered, and what evidence supports it, so a regulator, a Clinical Safety Officer or a procurement team can follow the logic and challenge it.
Fourth, we engage early. The people who will judge the approach should see it before it is final. Early, open conversations turn a potential rejection into a shared problem to solve.
Finally, we hold it to the same rigour as everything else. Novel does not mean lighter. If anything, the hazard analysis, the controls and the monitoring plan need to be clearer, because there is less precedent to lean on.
When the letter of a standard does not fit, its purpose still does.
A first, in practice
One example shows how this works. A customer’s product had been assessed by regulators as not being a medical device. On the surface, that looked reasonable. We asked them to look closer.
The product played a part in patient identification: making sure the right information reaches the right person. In clinical safety, that is one of the most critical areas there is. When identification fails, the wrong record, result or decision can reach a patient, and wrong-patient errors are a long-recognised source of serious incidents. We set out the reasoning plainly, with the hazards that would follow if identification went wrong and the controls needed to prevent them.
There were no counter-examples. The regulators could not help but agree. The outcome was a new category, one that navigated the fine line of classification as a medical device in a way that had no prior definition. It gave the customer a clear, defensible route forward, and it gave the system a reference point for every similar product that follows.
Why firsts matter
When a regulator or an NHS body accepts a new approach for the first time, it does more than unblock one product. It creates a reference point. The next team with a similar idea has something to point to, and the system learns a little faster.
That matters more than ever. The National Commission into the Regulation of AI in Healthcare has called for staged authorisation, earlier written feedback from the MHRA, and regulation that keeps pace with how products change. NHS England is reviewing DCB 0129 and DCB 0160 with AI and agile development firmly in view. The system is asking for exactly the kind of careful, well-evidenced thinking that novel products need.
If you are building something new
- Write down your intended purpose in plain language before anything else. Most classification questions start here.
- Map who is accountable for each clinical decision the product touches, across every organisation involved.
- Identify where your model does not fit the standard, and what the standard is trying to protect in each case.
- Document your reasoning as you go. It will be your strongest asset in every conversation that follows.
- Talk to the people who will judge your approach early, not at the end.
This is the part of our work that reflects our values most clearly: learning, doing the right thing, accountability, rigour and respect. New territory needs all five.