Fraunhofer podcast: Introducing AI, but how? Insights from KI-ULTRA

On 14 October 2026, the episode “KI einführen, aber wie? Das Projekt KI-ULTRA” (Introducing AI, but how? The KI-ULTRA project) was released in “Bridging Innovation”, the podcast of the Fraunhofer-Verbund Innovationsforschung. Host Constantia Bernhardt talks with Dr. Jan-Paul Leuteritz, who led KI-ULTRA at Fraunhofer IAO, and with Cedric Oette, Managing Director of Hawk Intelligent Technologies GmbH. The episode is about 48 minutes long and in German.

What KI-ULTRA was about

In the KI-ULTRA project, Fraunhofer IAO and the cooperating Institute of Human Factors and Technology Management (IAT) at the University of Stuttgart supported around 30 organizations in introducing AI, from micro-enterprises to DAX-listed corporations. Hawk was one of these companies.

The project produced two guidelines, one on strategy and change and one on carrying out specific AI projects. They are also available in English and French. In addition, an Evaluation Toolkit lets companies assess the conditions for using AI and the impact of an AI project they have introduced.

Hawk’s experience

With KI-ULTRA, Hawk wanted to tackle an initiative with a lot of leverage: decisions on which projects get priority in the company were to rely more on derived key figures than on personal preferences. There was no proven model for this initiative, which involved many stakeholders. The scientific support therefore offered a methodical approach in which a failure also counts as a finding.

Cedric Oette compares AI to the wheel, which on its own does little to help transport stones. With a board on top, it becomes a wheelbarrow, and in a well-designed construction, a conveyor belt. The benefit therefore comes from the overall construct in which the technology is embedded, not from introducing AI for its own sake.

Above all, the exchange gave Hawk insights for collaboration within the team. The researchers held up a mirror to the company’s own way of thinking, and the reports of the other companies showed how differently AI is introduced. Even in an AI company, not everyone is equally open to the technology. That is why it pays to discuss concerns openly in the team and to examine them seriously.

Technically, the deep learning approach chosen at the time would have required a great deal of preparatory work. Today, large language models shorten many of these paths because they can interpret human language well. Hawk continued the initiative after the research project and today considers it successful, also measured against the guidelines. In Cedric Oette’s assessment, what is often underestimated is the depth of integration needed with existing systems and data governance that can support the handling of sensitive HR and financial data.

Involving employees

Jan-Paul Leuteritz describes how differently the participating companies proceeded. There is no one-size-fits-all approach. Both guests advise against introducing AI without a clear goal. The first step, they say, is to clarify how a company earns its money and which tasks AI can sensibly take on there. According to Jan-Paul Leuteritz, co-determination and a say for employees make companies considerably more successful in implementation, because employees are the experts for their own workplace.

Listen to the episode

The episode, with show notes and transcript, is available on the podcast page (opens in a new tab) as well as on Spotify, Apple Podcasts and other platforms. The project page and guidelines are available from Fraunhofer IAO (opens in a new tab). Read more about the project and Hawk’s contribution in the knowledge article KI-ULTRA research project.