The SME Digital Center WertNetzWerke aims to make small and medium-sized enterprises (SMEs) fit for the digital future and is funded by the Federal Ministry for Economic Affairs and Climate Protection (BMWK). The Leipzig Value Creation Laboratory is based in Leipzig, in the premises of the Fraunhofer Center for International Management and Knowledge Economics IMW.
At the Life Science Stammtisch – “AI in Biotechnology” – on June 18, Sascha Mühl gave a lecture on the topic of “Qualification as an AI Guide for Bioscience, Medtec & Healthcare”. Sascha Mühl is a research associate at Fraunhofer IMW and project manager for AI and digitalization projects. So it was only natural that we asked him for a conversation on the topic “Why do small and medium-sized companies need this AI guide?”
Mr. Muehl, ValueNetworks is there for SMEs, at what size of company does it stop?
Small and medium-sized companies, that means we focus on the vast majority (99.4 percent) of companies in Germany, excluding corporations. This has something to do with turnover and the number of employees.
This is exactly the landscape that faces the greatest challenges with digitalization and AI. They can hardly afford their own IT department, let alone an AI department.
Exactly, no craftsman will be able to develop an AI solution himself, because as a craftsman he has other things to do.
You presented the AI Guide at the biosaxony regulars’ table, which can mean a variety of things. What is this AI Guide?
We want to pick out people in companies, employees or managers, depending on who is sent, and send them back to the company qualified as AI guides. With the aim of further advancing artificial intelligence in the company. We don’t want algorithm experts, i.e. specialists, but we want to train generalists.
AI Guide. Source: Mittelstand-Digital WertNetzWerke
If I understand correctly: You do not assume that these people have to understand AI in the truest sense of the word, i.e. how the algorithms work, but you want to enable them to work with existing AI solutions in the company?
So we are building on three pillars: AI management, AI technology and AI-related topics. AI management should enable the company to recognize the process: We now have an AI topic here, which can be at a very deep level, what do we do now, can we use ChatGPT or Gemini for this, for example?
Or you can put it at a very high level: We have a sales structure and we want a Sentiment analysis, an AI-based analysis of customer satisfaction, based on data that has yet to be collected. This requires a very deep understanding, a technological understanding. ChatGPT or Gemini, I would say, requires more of an understanding of related topics.
What about data protection, how do I handle the data that I send to America? We want to address these issues and make the participants aware of them.
Training with Sascha Mühl. Photo: Lichtwerke Design Fotografie
I have seen the Course takes place on seven daysas a qualification workshop. Is that sufficient?
It is a fairly compact course; other training courses last two or three years. We make it more compact because we don’t want to train super-experts. As I said, we don’t want to train specialists; we want to train generalists. There are many levels to consider in companies and this topic. It starts with the understanding of roles: what do I need in the company, who do I need to talk to, who do I need to pick up? – such as the data protection officer if there is an IT department, the IT department for server infrastructure and so on.
But then it is also important to have an understanding of technology. To understand what level am I at now? It is also important to understand where the focus lies, i.e. where should it actually go? This means that the first thing to do is to assess: do we even need AI for this? Do we have to develop the AI ourselves or are there solutions on the market?
This raises the question: Is an AI solution even suitable for my task?
I often hear requests like: We have an Excel workflow and want to try to solve it with AI. First of all, we need to capture it in the company in order to enable people internally to understand how to move forward. And then we also need to understand: We want to do something now, we have an internal use case here, i.e. practical approach that is economically worthwhile for us. And we also know that we need external help, so the AI guide knows who to contact.
To understand the technology, we start at the starting point: what technologies are there and how do they work in general in the company’s processes. We experiment a bit with them so that we can later approach specific experts. There are many AI experts, and wading through this jungle is difficult. Our industry focus is health bioscience, which means we are going into the medical field, which is already quite experienced with AI per se. At least those who develop products.
That will be the largest target group. Those who first want to learn the basics of how to deal with AI in the company and those who start at the highest level and see AI as a product. We differentiate between the groups of participants: I want to optimize my processes with AI and I want to prepare an interview with ChatGPT.
Now we’ve covered this in broad terms, we can’t go into all the details anyway. Using AI also requires a certain understanding of legal processes, you already mentioned data protection. If I work with AI, do I also need to know whether I’m on legally safe ground? You mentioned this in your presentation, you don’t train lawyers, but you tell the participants what they need to be aware of?
We are opening levels, that is the only thing that can be done at the moment because the legal situation is not yet clear. Recently, the EU-AI Act has become legally binding and now attempts are being made at a high political level to break the whole thing down until norms and standards emerge at some point. There are many different levels to consider, it’s not just cyber security, it’s also about the security of AI models. For example, how likely is it that I can manipulate my own language models, similar to ChatGPT?
This legal framework has by no means been translated into binding standards. In the medical field, the standards are already quite advanced due to long experience or rather long establishment. What will change as a result of the EU AI Act is, for example, the basis of risk assessment. The implementation of AI in medical devices will have to be reassessed by companies.
Especially in research, when I go out and develop something, I have to at least know that I shouldn’t type everything into ChatGPT publicly. As I understand it, it’s about making people aware of the legal issues here. What is certain, what is still open?
What is important to note and what you should definitely know is that AI makes decisions based on probabilities. An imaging procedure for analyzing abnormal areas in the body, such as that used for years in magnetic resonance imaging (MRI), makes decisions based on probabilities.
And the question arises: Is it legally safe to use this imaging procedure in order to then perhaps not be able to rule out a certain type of cancer that the patient might still die from? This question already exists, but it will have to be reassessed in connection with AI. In what form, no one can tell you in detail at the moment.
You are dealing with the topic in the life sciences area, let’s call it that, this AI guide is probably also suitable for other areas.
That’s easier to answer in retrospect. The Mittelstand-Digital Zentrum WertNetzWerke has various partners who have already tested an AI guide. That was more in the industrial sector. That means there were production specialists involved, but there were also sales people involved and it went quite well.
We then decided to take the three sectors of medical technology, bioscience and health management due to our research focus in the “Digital Health” group at Fraunhofer IMW, where there are other accompanying topics. There is the issue of law; product law in medicine is a completely different issue to product liability for standard products of a manufacturing company. That’s why we took this concept and are now trying to apply it specifically to these sectors.
But this open concept still exists?
This open concept exists. In 2023, it went quite well. The colleagues in the Brandenburg Region and in North Rhine-Westphalia, they had such high demand that they are now offering it again in August.
Because these are industrial regions and former coal mining areas (Lausitz, Central Germany, Rhineland), we are again trying to make the AI Guide future-oriented and thematically oriented for these classic producers. We are trying to transfer this to another scientific field and another industry segment with the life sciences industry.
To summarize: The employees who are trained as AI guides are then able to organize the introduction and operation of AI in the company. Not the technical side, so they don’t have to program algorithms, nor will they provide legal advice. But they know where the control points are where something needs to be done and they have an understanding of what is actually possible.
Yes, that sums it up well. The effort involved is immense. I am currently also leading a project in the leather and textile industry at the SME Digital Center WertNetzWerke, Leipzig Value Creation Laboratory. This expectation also prevails there: ChatGPT has arrived, and everything that was not possible before can now be done with AI. I will specifically clear up this hype in the first workshop. Here I will raise awareness of what can realistically be achieved with the individual AI technologies at the moment.
It would be far too much work for a company if they had to send someone externally to some kind of training course. That’s why we want to create this liaison office, to be a point of contact for companies and employees, to filter out whether investing in AI solutions is worthwhile for small and medium-sized companies or not.
Mr. Mühl, thank you for the interview and your time.
