Increasing AI Use for Device Master Files
Artificial intelligence is booming. Large language models (LLMs) such as ChatGPT, Claude and Gemini have not only gained incredible capabilities in recent years, but are also enjoying ever-increasing popularity.
Whether in science, public media, private contexts or even regulatory affairs, support from digital text, image and audio generators has become indispensable.
The use of chatbots is currently on the rise, particularly in regulatory affairs. This is hardly surprising: the requirements of the MDR or individual standards often appear complex, confusing or simply bureaucratic and dry.
It is therefore useful when AI formats documents, compiles evidence or puts together entire Device Master Files. But be careful! Because electronic brains are not that ‘intelligent’ after all.
Human Intelligence Remains Paramount
AI systems are great for sifting through and summarising large amounts of data, or for assisting people in reviewing important documents.
However, they do not ‘think’. The widely used LLMs are based on statistical data analysis: they spit out what is most likely. They do not recognise errors because they are ‘wrong’, but because they are ‘statistically unusual’.
Commercial chatbots are also trained to produce ‘pleasing’ results. They generate outputs that feel correct – even if that may not always be the case. This is simply due to the nature of their internal structure and development.
That's why even the best AIs sometimes fail to recognise the simplest errors – a circumstance that has already driven many AI users to distraction, as some of you are surely aware.
Human-Machine Cooperation is the Best Choice
As scientific studies describe, mastery lies in complementing the skills of humans with those of AI.
This means that expert knowledge is more important than ever: the errors that AI can generate are often hidden and hardly noticeable. It is therefore essential that people with in-depth specialist knowledge check the AI results or formulate the prompts for the chatbot.
Particularly Critical in Regulatory Affairs:
Precisely because regulatory affairs demand precision, precisely because normative requirements can often be counterintuitive or confusing – especially where errors do not stand out but can creep in without being immediately noticeable – it is important that humans monitor the work of machines. This is particularly important because in regulatory affairs, small errors can have serious (and costly) consequences.
And in order to exercise effective control, experts are needed who can directly identify potential errors in AI.
Conclusion: AI! but with Caution
BEO BERLIN expressly supports AI-assisted work in regulatory affairs. However, our experts also recognise that AI systems cannot (yet) replace expert knowledge – AI-generated errors are noticed by the regulatory authorities at the latest.
We therefore recommend customised AI support. Always in the context of the capabilities already available in your company and the requirements of the processes to be optimised.
If you are looking for a consultation or need experts to ‘proofread’ the results of an AI, contact BEO BERLIN!
(3 Rating(s), average: 5.00 out of 5)
Loading...
