In the interview with Josef Dietl, founder and managing director of Pro Digi Par GmbH.

In a world where artificial intelligence (AI) is becoming increasingly prevalent, it is crucial to understand the reality behind the hype. But what is the actual state of AI adoption in companies? What role does it play in corporate structures, and how can it be used effectively? In an exclusive interview with Josef Dietl, we will delve into these very questions. The discussion will focus on clarity regarding one's own positioning and the challenges involved in implementing AI. After all, as the saying goes: AI cannot read minds, but it can support companies on their journey if used correctly.

Since its founding in 2019, Pro Digi Par GmbH has made a name for itself, particularly among small and medium-sized enterprises, successfully guiding them on their digitalization journey. As the “Prozess Digitalisierungs Partner,” the company supports its customers in overcoming growth barriers, defining corporate goals, and guiding them through the entire digitalization process and beyond. The motto “Freude am Fortschritt” (Joy in Progress) is at the heart of everything they do, with a particular focus on the harmonious interplay between humans and technology.

Josef Dietl studied physics at TU Munich to “understand what holds the world together at its core,” and even then worked on neural networks. A stint at CompuServe led him to the team of Sir Tim Berners-Lee, the inventor of the World Wide Web. There, he worked, among other things, on knowledge representation. Later, during his time at SAP AG, he acquired around a dozen patents, including some for classifiers that would likely now be considered AI technologies. To gain deeper insight into the vision and working methods of Pro Digi Par, we interviewed Josef Dietl and discussed the sometimes overestimated capabilities of AI and its nonetheless incredible potential for business applications.

DIGITAL FUTUREmag: Mr. Dietl, how do you generally assess the current hype surrounding artificial intelligence?

Josef Dietl: The other day, I came across a joke: A girl asks a parrot, “Can you talk?” and the parrot replies, “Yes, but I have no idea what I’m saying.” That’s exactly what AI does today. AI has made remarkable progress in recent years, but its capabilities are often dramatically overestimated. There’s no “anchor” in reality. Like the parrot, AI can generate beautiful texts, images, and videos—but it has no understanding of what it’s talking about. That’s probably why the most visible use case for AI right now is marketing: it’s the “industry of the creative.”

DIGITAL FUTUREmag: How important is it for companies to clearly define their own positioning and strategic direction before implementing AI?

Josef Dietl: This is precisely where companies need to make their own contribution: Even if it feels this way, AI cannot read minds. A clear company context helps, for example with positioning and strategic alignment.

DIGITAL FUTUREmag: What role does corporate structure play in the successful integration of AI?

Josef Dietl: Only when you know yourself well enough can you identify tasks that can be handed over to AI and prepare these tasks appropriately. This starts with strategy and seamlessly transitions into the organizational structure and processes, including organigrams and workflows. For AI to "function" in day-to-day business with its strengths and weaknesses, the processes for handing over tasks and the associated data to AI—and how the results must be further processed—need to be clarified.

DIGITAL FUTUREmag: What challenges do you typically see when implementing AI in companies, particularly regarding data quality?

Josef Dietl: AI still delivers results according to the principle of "garbage in, garbage out." To generate reliable results, you need to control the quality of the input data. If the input is correct, AI can deliver great results even from minimal information—but if the input is weak, AI can also completely miss the mark.

DIGITAL FUTUREmag: How can companies ensure that the data they use for AI applications is accurate and reliable?

Josef Dietl: Smaller companies in particular often need to take a step forward in their internal data quality. A few practical examples: orders, contracts, documentation, or accounting—especially in smaller companies, these depend on employee expertise. The results are correct, but the formats can vary, and repeatable processes often don’t exist; every document is created in a new way. This creates diversity without meaning, which is not optimal as a basis for AI.

DIGITAL FUTUREmag: What steps should companies take to benefit from AI in their core business?

Josef Dietl: For AI to work optimally in these cases, it’s useful to standardize processes and bring the relevant data into classic enterprise software like SAP or Odoo. This way, "classic IT," with its focus on data and process quality, complements the strengths of AI.

DIGITAL FUTUREmag: Can you give examples of companies that have successfully benefited from AI, and what measures did they take to achieve this?

Josef Dietl: AI has long been used in accounting, even long before the current hype. The type of AI used is called "classifiers"; OCR has been gaining importance for a long time, and lately, text comprehension has also come into play. After all, accounting processes have been standardized for centuries, which made it an obvious candidate even when AIs were still less powerful. To enable classifiers to deliver optimal benefits, the receipts are pre-processed and standardized so that they can be batch-processed by the AI. In the future, this process will become even more efficient with e-invoicing.

DIGITAL FUTUREmag: What role does the Pareto principle play in evaluating the benefits of AI in companies, and how can companies incorporate this into their strategy?

Josef Dietl: The example of accounting illustrates this very well: In the vast majority of cases, the AI delivers excellent results that can be used without further ado. Depending on the scenario, this works in 80%, 90% of cases or even more. And sometimes there are curious surprises whose identification and correction are labor-intensive and require specialists. For the time being, you therefore have to manually "proofread" the results and reserve capacity to handle the "remainder". This is how I would approach any AI project for the time being: Standardize the data inflow via normal company software such as Odoo, leave the Pareto share of the work to the AI and, at the same time, plan the necessary capacity and expertise to deal with the "remainder".

DIGITAL FUTUREmag: Many thanks for this truly exciting interview and the wide range of topics we were able to cover in this conversation.