Artificial intelligence is moving rapidly into healthcare.
Regulation is moving with it.
For MedTech companies, that creates a fundamental question:
Are you designing an AI product and then asking how to get it through regulation, or are you designing a product that can succeed within the healthcare system from day one?
There is an important difference.
The UK’s AI and Digital Regulations Service is helping to make that regulatory landscape clearer. But it also highlights something every MedTech development team should recognise.
Bringing an AI or digital health technology to market isn’t one regulatory problem.
It is a system of interconnected ones.
Welcome to the regulatory maze
The AI and Digital Regulations Service, or AIDRS, brings together guidance from four major UK bodies:
NICE, covering evidence, clinical and cost effectiveness.
MHRA, covering medical device regulation.
HRA, covering health and social care research.
CQC, covering the regulation of health and social care services in England.
Commissioned by NHS England, the service provides developers and adopters with a centralised source of guidance for navigating AI and digital health regulation. [1]
That is important because the journey can become complicated surprisingly quickly.
Is your software a medical device?
What classification does it fall under?
What evidence will NICE expect?
Will you need clinical research?
How will patient data be handled?
Does the technology meet NHS clinical safety requirements?
What about cybersecurity?
Interoperability?
Accessibility?
Human factors?
Algorithm updates?
Post-market monitoring?
And perhaps the most commercially important question:
Who will actually adopt it once it is approved?
The recently published National Commission into the Regulation of AI in Healthcare acknowledged this complexity. It noted that developers may need to navigate requirements involving MHRA, NICE, HRA, CQC, the ICO and other bodies, alongside medical device regulation, evidence generation, health technology assessment, NHS assurance and data governance. [2]
For innovators, particularly smaller businesses and teams developing something genuinely new, that is a substantial challenge.
Regulatory strategy is product strategy
There is a temptation to treat regulatory work as a gate towards the end of development.
Build the technology.
Prove it works.
Then work out how to get it approved.
For AI and digital health, that approach can be expensive.
Imagine spending 18 months developing an algorithm before discovering that the evidence being generated doesn’t answer the questions required for adoption.
Or building an elegant clinical interface that doesn’t fit NHS workflows.
Or designing an AI tool around data that cannot practically be accessed at scale.
Or proving algorithmic accuracy without proving that the technology improves outcomes in real clinical environments.
NHS England’s experience evaluating AI technologies illustrates the breadth of the challenge. Real-world evaluations from the AI in Health and Care Award considered safety, accuracy, effectiveness, value and fit with healthcare sites, including whether technologies were acceptable to clinicians and patients. [3]
So perhaps the better question isn’t:
“How do we get this product through regulation?”
It is:
“What product should we develop if we want it to be approvable, adoptable and commercially viable?”
That changes the development conversation considerably.
Compliance doesn’t guarantee adoption
This is where things become particularly interesting.
A product can be technically compliant and still fail commercially.
NHS England’s Digital Technology Assessment Criteria, DTAC, considers areas including clinical safety, data protection, technical security, interoperability, usability and accessibility. [4]
All are essential.
But passing requirements doesn’t necessarily mean clinicians will change behaviour.
It doesn’t mean patients will trust the technology.
It doesn’t mean the workflow works.
It doesn’t mean a procurement team sees sufficient value.
And it doesn’t mean someone has a budget to buy it.
So when should adoption be designed?
At Maddison, we would argue: from the beginning.
Regulation, human factors, clinical workflow, usability, evidence generation and commercial strategy shouldn’t be separate workstreams that eventually collide.
They should inform one another throughout development.
AI creates a different development challenge
Traditional medical devices can often be defined relatively clearly.
AI can be different.
Algorithms evolve. Data changes. performance may differ between populations. Models may be updated. Clinical workflows change around them.
The MHRA’s AI Airlock programme is exploring precisely these challenges. In 2026, work within the programme examined approaches to allowing AI medical devices to be updated more frequently while maintaining appropriate regulatory control and patient safety. [5]
That raises some difficult product development questions.
What happens when your product changes after approval?
How do you validate an algorithm that continues to evolve?
Which changes require further regulatory consideration?
How do clinicians understand what the AI is doing?
How much explanation does the patient need?
What happens when the algorithm is wrong?
And critically:
Who is responsible for the decision?
These aren’t questions to ask shortly before launch.
They can influence the architecture of the product itself.
This should change how innovation teams work
Large MedTech organisations are extremely good at developing products through established processes.
But genuinely new AI opportunities can sit awkwardly within those processes.
There may be too many unknowns.
The clinical proposition is still evolving.
The regulatory pathway isn’t completely defined.
The technology needs experimentation.
The workflow needs testing.
The commercial model may not yet be clear.
This is where a small, multidisciplinary development team can be valuable.
Take the opportunity away from the constraints of the core development programme.
Bring together industrial design, UX, human factors, engineering, regulatory thinking and commercial strategy.
Then interrogate it.
Does the technology solve a meaningful clinical problem?
Can we build it?
Can we regulate it?
Can clinicians actually use it?
Can we generate the evidence required?
Can healthcare systems adopt it?
Will somebody pay for it?
Build enough to answer those questions.
Prototype.
Test.
Learn.
Challenge assumptions.
Kill weak ideas early.
Accelerate the strong ones.
Then make the larger investment decision with evidence rather than enthusiasm.
Regulation can become an innovation tool
The AI and Digital Regulations Service is useful because it brings much of a fragmented landscape into one place.
But there is a bigger lesson here for MedTech leaders.
Regulation doesn’t have to be something that slows innovation down.
Used early enough, it can help define what good innovation looks like.
It can expose weak assumptions before millions are committed to development.
It can influence architecture.
It can shape clinical evidence.
It can improve usability.
And it can force development teams to confront the question that ultimately matters:
Does this technology have a realistic pathway from clever idea to everyday healthcare?
At Maddison, we believe the strongest AI and digital health products will emerge when design, engineering, human factors, regulatory strategy and commercial thinking happen together, not sequentially.
Because getting an AI medical technology approved is only one milestone.
Getting it used is what creates value.
References
- AI and Digital Regulations Service, Understanding regulations of AI and digital technology in health and social care, NICE, MHRA, HRA and CQC, 2026.
- Department of Health and Social Care, National Commission into the Regulation of AI in Healthcare: Recommendations for a future regulatory framework, September 2026.
- NHS England, Planning and implementing real-world artificial intelligence evaluations: lessons from the AI in Health and Care Award.
- NHS England, Digital Technology Assessment Criteria and Digital Clinical Safety Strategy.
- AI and Digital Regulations Service, Exploring the future of Medical Device regulation with the MHRA’s AI Airlock, July 2026.