AI Help in the Shadows
E3: Mr. Failer, everyone is talking about artificial intelligence at the moment. However, only larger providers seem to be benefiting from it. Have you been missing out?
Thomas Failer: You're right. Artificial intelligence is currently a dominant topic. However, the discussion here is always about generative AI, which indeed opens up completely new application possibilities. However, AI isn’t something completely new. AI has been used for some time now, particularly in the form of machine learning. For image diagnosis options in medicine or predictive maintenance in mechanical and environmental engineering, for example. Incidentally, previous AI and the new generative AI have one fundamental thing in common: they need to train with a good deal of data in order to achieve good results.
E3: You have a lot to do with data, that's your core competence. What will change for you with the new generative AI?
Thomas Failer: The question is rather: what will change for customers and what are the specific benefits for businesses? We always think from the customer's perspective. And that's why we've been working with AI for a long time, even before the topic caused a furor, but until now it has been more in the form of machine learning. Building on this, however, we’re currently also working on the use of generative AI. For this very reason, we set up a dedicated agile development team for AI back in 2022, which builds on our previous activities in this area and drives them forward.
E3: So it wasn't ChatGPT that brought DMI closer to the topic of AI?
Thomas Failer: Yes and no are both the right answer. Yes, because it provides new opportunities for us. No, because we’d already been working on it before, but we simply chose not to shout it from the rooftops.
E3: Why did you work on it earlier?
Thomas Failer: AI is not about intelligence as an end in itself. From a corporate perspective, AI is always a means to other ends. It may be interesting for researchers, the press, or science fiction authors, but company leaders in companies are not there to dream, but rather to plan. How can we become more efficient? How can we relieve employees of routine tasks? How can we speed up IT projects? From a business perspective, these are the key questions for every technology, including artificial intelligence. To put it another way, AI is always about its potential for automation in a company. You could even say that automation is a synonym for AI.
E3: Then please explain to our readers where the development focus in terms of AI has been at DMI so far and why you are starting to talk about AI now of all times.
Thomas Failer: We specialize in managing the lifecycle of data lifecycle, especially legacy data from SAP and non-SAP systems. In this context, there are two scenarios for which our offering and AI are essential. Firstly, the SAP transformation, which from the customer's point of view is literally about not losing the race against time but winning it in a brilliant final spurt by 2030, at the latest. With our approach, it’s already possible to halve the time required for this. However, our goal is to reduce the effort required by the same amount in the coming years. And that's where AI comes into play, of course.
E3: You mentioned two scenarios - which is the second?
Thomas Failer: The topic of retention management is quite prevalent and is an ideal use case for AI. Since September 1, a new data protection law has also been in force in Switzerland, which essentially adopts the provisions of the European General Data Protection Regulation. This has been another wake-up call for many companies as it i’s an immense challenge to find personal data in all systems, applications, and archives, and to delete specific data if necessary.
E3: How can AI help here?
Thomas Failer: We have the impression that most efforts in this field have so far been limited to data, even if this structured information in the form of metadata relates to unstructured information, such as documents or images. A pragmatic approach has looked something like this: personal data should is be found in such and such tables and fields if everything has been properly maintained and stored. For example, the first name and surname is in a master data record or in the description of a person’s photo. And then the algorithms and rules for retention management were only applied to these tables and fields. But what if this data was also stored in other places or wasn’t stored on an image database at all, for instance? Here, AI can help identify and adapt the algorithms.
E3: How exactly?
Thomas Failer: Thomas Failer: Essentially through machine learning. The AI can be trained using e-mail messages, for example. This enables it to recognize names, even if their spelling in an e-mail address differs from normal spelling rules and only appears in the copy address field, but not in the message text itself. Or it can use the content to recognize that there are first names in column X of this or that sub-table or Z-table, even if the description of the column or table doesn’t suggest anything in this regard, simply because the AI knows most of the first names used worldwide. AI can also be used to recognize content such as International Bank Account Numbers (IBANs) or social security numbers, allowing conclusions to be drawn about the people behind these numbers. And let's not forget all the personal data that’s available as unstructured information, such as in documents, notes, and logs, and is stored as PDF files or in other file formats. Identifying the relevant data alone is a huge area where people require AI support.
E3: Thank you for the interview.