Data and AI: the crucial question

So the question is: how can AI's promise of productivity be fulfilled without compromising the necessary data sovereignty? Thomas Failer, founder and Group CEO of Swiss company Data Migration International, spoke to E3 magazine about this.

E3: Quiz question: Do you know what the Gretchen question is and where it comes from?

Failer: Klingt nach Deutschunterricht. Im Zweifel lautet die Antwort „Goethe“.

E3: That's right! That is Gretchen's question to Faust about how he feels about religion. The name for a crucial question has developed from this. So here is the central question for 2026 and beyond: What do you think about AI?

Failer: We use AI in our work and are incorporating more and more AI-based functionalities into our platform.

E3: Could you please give a few examples of this?

Failer: With pleasure. You will be surprised to hear that we have been very busy over the past year. There is the intelligent Business Object Proposer, BOP, the Personal Data Identification, PDI, functionality, an AI chatbot that makes our platform talk, so to speak, an intelligent low-code development for transformation projects, LCT, and an AI functionality for Data Quality Improvement, DQI.

E3: Could you please explain in a few words what these innovations do and why you developed them?

Failer: One of the central challenges of transformation projects to SAP S/4 Hana is the development of specific business objects. Our Business Object Proposer takes on the necessary tasks to find the correct and complete data in the right tables and link them together correctly. These tasks are tedious and time-consuming. Automating them with the help of our BOP has the potential to halve the development time of business objects.

E3: Personal Data Identification - isn't personal data old hat?

Failer: Quite the opposite. You'd be surprised how many organizations continue to struggle to find and manage personal data reliably and in full. Our AI-based PDI functionality searches all systems, especially those with no or poorly maintained metadata repositories, and classifies tables, columns and fields. It then presents the results of its analysis and classification for review. This gives companies an overview of where personal data is located and they only need to fine-tune the proposed results. This reduces the search effort to a minimum.

E3: Touché - and the chatbot really makes your platform talk?

Failer: That is indeed the case. Research is a key discipline for so-called large language models. This is precisely why we have trained an AI for this use case and integrated it into JiVS IMP as a chatbot functionality. This makes it possible for employees in a company's finance department, for example, to search for and find data and documents that have been historicized on our platform for legally compliant long-term storage using natural language input, such as: „JiVS IMP, I have a problem. I need to collect the open items for customer 1020 and the 2019 financial year. Unfortunately, I am not familiar with the individual development in which the documents were created. Could you please find them?“ In response, the chatbot on our platform would display the documents searched for, sorted by hit probability, and ask the user to confirm that it is the information they are looking for.

E3: Let's move on to the low-code transformation - what is it?

Failer: In transformation projects, it is the specialist departments that write the necessary specifications alongside their actual work so that IT can create the appropriate transformation rules. The effort involved is enormous. Every single transformation rule has to be programmed. Writing the program code is less important than the quality of the results achieved. This is where LCT comes into play. The intelligent functionality automatically converts the input in natural language into MS SQL statements, one of the most widely used coding standards in the world. The advantage: MS SQL has not only always been the JiVS-IMP programming language. In fact, it is supported by the vast majority of AI models due to its close proximity to natural language.

E3: That leaves the issue of

Data Quality Improvement.

Failer: Transformation projects are actually the ideal opportunity to clean up your own database. The vast majority of companies shy away from the effort involved, which is understandable. But with our platform, this hesitation can no longer be justified. JiVS IMP automatically sorts out duplicates, reducing potential errors by 75 percent or more. Together with the usual reduction of 90 to 95 percent of transaction data and 50 percent of master data during the transformation to SAP S/4 Hana, the effort required to optimize data quality is already massively reduced. But it gets even better: with the help of Data Quality Improvement, the remaining effort for data quality assurance can be further reduced considerably, in our experience by half or more. And this brings us closer to the real crucial question.

E3: And what is it?

Failer: Wie halten Sie es mit den Daten? Das ist die entscheidende Frage, die unsere Kunden beim Thema KI zurzeit umtreibt.

E3: Why is that?

Failer: Companies have learned from experience. Too many AI projects have failed. And the key reason for this failure is the lack of quality, availability, accessibility and independence of the data.

Wie halten Sie es mit den Daten? Das ist die entscheidende Frage, die unsere Kunden beim Thema KI zurzeit umtreibt.

Thomas Failer,
Founder and Group CEO of
Swiss Data Migration International

E3: Why do we hear so little about it?

Failer: It always takes a while for such findings to take root. From the very beginning, the discussion has mainly revolved around algorithms, models and computing power. Not that these aspects are not important. But in the business context, the fuel for the models is even more important: and that is the company-specific data.

E3: It's no secret that they are not always in the best condition.

Failer: You are right. Data is often fragmented, duplicated and triplicated, incompletely managed and locked up in outdated systems.

E3: Why do you think that is?

Failer: Application and system landscapes have grown historically. As a result, complexity and susceptibility to errors are increasing. This is not the result of ignorance or negligence. IT managers are well aware of this problem.

E3: So why don't they do anything about it?

Failer: They want to and are doing a lot to achieve this. But the main problem, apart from the usual workload, is the fact that data is too intertwined and wired with applications and systems. Data is only valuable in the right business context. And this can usually only be created in and with the applications in which the data was created and processed.

E3: This is nothing new either. So why is the problem so acute in the AI context?

Failer: AI scales the problem. If the foundation is fragile, cracks will form in the walls after a short time, especially in the load-bearing parts, meaning the house could collapse at any time. Users must be able to trust the results of the AI. How is this possible if the database is fragile? Companies cannot check every result at great expense before using it to support their employees and fully or partially automate processes. If this were necessary, the efficiency and productivity gains would be wasted.

E3: And companies are just becoming aware of this?

Failer: Absolutely. We noticed this again in January at our Digital Lounge@Davos event (see text box on page 36). A pattern emerged in the discussions between managers and AI specialists, as well as in the hackathon that we organized in parallel to transform AI concepts into functioning solutions: The limiting factor was rarely the algorithm. The decisive factor was whether data was available, comprehensible and decoupled from legacy applications.

E3: What is your conclusion?

Failer: If data is correct, subject to proper governance and accessible independently of its applications and systems, the intended innovations can be implemented quickly. Data independence is therefore becoming a strategic and central capability for successful AI projects.

E3: Thank you for the interview.