Automation is Advancing.
The medical technology industry is one of the most innovative industries in the world. New inventions are the order of the day here, and it is not uncommon to imagine manufacturers of medical devices as real-life Gyro Gearlooses.
It is therefore not surprising that artificial intelligence was already present in the medtech industry long before the mass phenomenon of large language models (LLMs).
What is new, however, is that AI is now increasingly being used in regulatory affairs. Not as a replacement for experts, but as a tool that – when used correctly – can make processes faster and more consistent. And this is precisely where BEO BERLIN provides support: individually, pragmatically and with a focus on what really matters in everyday life.
AI has Long been Highly Specialised
Interestingly, LLMs did not initially play as significant a role in medical technology as they did in other fields. In the early 2020s, they were simply too unreliable for a precise and highly regulated environment, while established machine learning systems – especially in image analysis – had already found their niches.
Broader Expertise – Broader Application
Gerade aus Präzision und Regulation folgt aber die Notwendigkeit, mit großen und diversen Datenmengen umzugehen: technische Dokumentationen, QM-Prozesse, PMS-Daten, Tabellen, Freitexte, Bilder, Codes müssen gemeinsam interpretiert und verarbeitet werden.
This is where the realistic added value of multimodal AI approaches lies – often as a combination of LLMs and specialised tools. Not because an LLM can ‘do everything’, but because it can provide targeted relief: sorting, clustering, comparing, summarising, flagging anomalies. The benefit does not come from magic, but from clean process integration.
Broader Application – Higher Error Rate
But beware! LLMs can formulate convincing statements, but they do not possess genuine understanding. This increases the classic risks: fabricated claims, incorrect weightings, statistically plausible nonsense.
In short: The broader the application of AI, the more important it is for qualified individuals to monitor its outputs.
So Where is the Relevance for Manufacturers?!
Thank you for reading this far. Admittedly, the connection to regulatory affairs has been rather indirect so far.
That is set to change.
This year, BEO BERLIN received several enquiries from manufacturers who already use AI to some extent for technical documentation or internal quality management – or who wanted to examine the extent to which AI support is currently feasible in regulatory affairs.
It is clear that the authorities are taking this development seriously: The Norwegian DMP demands transparency on how AI was used in documentation packages – and makes it clear that responsibility remains entirely with the submitter.
The FDA is also pushing ahead with its use of AI – especially internally. In its announcement on 1 December 2025, it announced that it would start using AI to evaluate applications with immediate effect. This will be done under human supervision, but will be much more autonomous than before, including for pre-market reviews, review validation and post-market issues.
And in Europe, the AI Act is being rolled out in stages. Key obligations will take effect on 2 August 2026; high-risk rules for regulated products will follow according to the EU timeline from 2 August 2027.
MedTech Europe is therefore calling for consistent interaction with MDR/IVDR and a postponement of the application for AI systems affected by MDR/IVDR until 2 August 2029 in order to avoid double regulation and delays.
What Does it all Mean?
There is no question that sorting and structuring large amounts of data using AI can be extremely helpful for technical documentation or post-market surveillance. This makes it all the more important not to lose sight of the weaknesses of generative AI.
BEO BERLIN therefore recommends: AI-supported solutions should always be double-checked by experts. Not AI instead of expertise, but AI with expertise and clear roles lead to clean documentation.
However, properly integrated AI support is likely to become a real competitive factor in the medtech industry by the end of the decade. The efficiency gains achieved through well-designed AI-human interfaces are real – as is the danger posed by AI-generated nonsense when control logic and responsibilities are lacking.
If you would like assistance in preparing an application for approval, setting up or improving a QM system, or require reliable market surveillance, please contact BEO BERLIN.
Together with you, we evaluate individually and pragmatically which AI solutions make sense for your project – and where the best compromise between efficiency, security and regulatory resilience can be found.
(1 Rating(s), average: 5.00 out of 5)
Loading...
