Great interest, many questions

Arno Engel is actually at home in the world of finance. As finance officer at the state representation and the Ministry of Finance of Baden-Württemberg, he deals with budgetary issues day in, day out. But on the side, he is building a network across departments to promote the use of artificial intelligence. ‘We are an organisation with around 400 people, we don't have any data scientists, so there was no one who could take the lead on this topic,’ Engel reported at SCCON 2025. ‘Our approach was to look for people from all departments, regardless of whether they had any expertise in the field or not.’

Training courses with real-life use cases

A highly committed team has now been formed, whose suggestions are well received because they come from the ranks of the employees themselves rather than being imposed top-down by experts. For example, they developed internal training courses themselves because it is of little use to administrative staff if external trainers explain to them for the umpteenth time how to shorten texts using large language programmes.

‘These training courses were often far removed from the everyday work of the employees,’ said Engel. So the volunteer team first asked around to find out where the problem lay and discovered that There was great interest in working with AI, but many were unsure what they were actually allowed to do. Some were keen to experiment, while others work with sensitive data and are therefore particularly cautious. Now there are internal training courses for a general overview and workshops with real-life use cases from within the organisation. However, when it comes to procuring AI solutions, the team relies on the expertise of the Innolab in the State Ministry.

Particularly complex with AI: buy or develop in-house?

The procurement of AI solutions also poses questions for the experts themselves, above all: develop in-house or buy? This decision has become much more complex, says Johannes Neumeier, Head of AI and Deputy Head of the Cloud, Platforms and Data Management Unit. In the past, SAS solutions were purchased as turnkey products. With AI, however, every layer now raises the question: build it yourself to maintain control, or buy it to reach your goal faster and benefit from external expertise?

Neumeier said he would be in favour of doing both and advocated taking the plunge: As soon as there is clear data governance and guidelines, he would be in favour of using a team's coding skills to drive solutions forward. It is important that ‘it is open source and that it can be reused’ – so that everyone can learn from each other. When it comes to procurement, he would like to see certain specifications for the architecture of an application that ensure compatibility with other solutions.

Added value requires commitment

Juliane Braun, Chief Data Scientist at the Federal Ministry of the Interior, who is promoting AI applications, agreed. She believes that the real added value of AI lies in constant exchange and close cooperation across departments, but also with colleagues who do the technical work. Open data is also part of this, because although summarising texts is nice, ‘we generate real added value when we start collecting official data’. This requires ‘a very clear commitment on the part of the administrations that we will work on these issues together’.