Guest article by AKDB: AI and data protection are not mutually exclusive
AI can speed up administrative processes and reduce the workload on staff. Secure infrastructure, clear rules and the transition from pilot projects to routine operation are crucial.

AKDB at the Smart Country Convention 2025. Photo: Messe Berlin
AI reduces the administrative burden on staff, makes administrative knowledge more easily accessible and improves public services. However, the prerequisites for secure and sustainable operation are often lacking. Christian Davepon, Head of AI at AKDB, discusses which solutions already exist today and what can be expected in terms of AI in five years’ time.
Where can AI already make a measurable difference in public administration today, and in which use cases is its potential still being overestimated?
AI is currently making a difference in areas where staff spend a great deal of time each day searching for information, processing documents and carrying out repetitive tasks. It saves local authorities an enormous amount of time when knowledge is made accessible more quickly.
There is also great potential in the semi-automated, AI-supported checking of application documents for completeness and plausibility. With our Intelligent Administrative Assistant iva, we are already gaining practical experience with this in immigration offices and housing benefit offices. In the case of housing benefit, pilot local authorities report time savings of up to 40 per cent per case.
AI is overestimated in situations where it is expected to compensate for unclear processes or a poor data foundation, or to make decisions entirely automatically. Particularly when it comes to legally relevant decisions and matters of discretion, professional responsibility must remain with administrative staff.
Many public authorities are experimenting with AI, but far fewer solutions make it into routine operation. In your view, what are the biggest hurdles between a pilot project and widespread implementation?
For individual local authorities and IT service providers, it is hardly cost-effective to set up their own AI applications, operational concepts and operational teams. Added to this is the effort involved in procurement. From our discussions with local authorities, we know how much time is taken up by selecting suitable providers and reviewing data protection, compliance and IT security.
Local authorities therefore need reliable standards, transparent procedures and professional operations. To this end, AKDB has taken the lead in developing the GenAI platform on behalf of govdigital, the cooperative of public IT service providers. It offers participating IT service providers a shared environment for the development and operation of AI applications. This allows applications to be reused without each provider having to set up its own infrastructure. The platform is going live here at the Smart Country Convention.
How exactly is AI changing the work of local government staff, and which skills, processes or structures need to change so that it actually reduces their workload rather than creating additional work?
AI only creates breathing space if it is integrated into existing workflows.
The professional expertise of staff does not become less important as a result, but rather more so. They must be able to interpret results, identify errors and decide when AI may be used. This requires training and clear rules.
I also expect a shift in operations: away from numerous individual on-premises solutions, towards jointly operated cloud platforms.
Data protection and transparency are particularly sensitive issues. How can AI be used in such a way that trust and the requirements of the rule of law do not become obstacles to innovation – but rather form part of a sound solution?
Data protection and transparency are not at odds with innovation. They are the prerequisite for an AI application to be used in public administration on a long-term basis at all.
With KAI, our AI assistant for local authorities, we are putting this into practice.
KAI is operated on govdigital’s GenAI platform within the public sector. Each local authority is provided with its own, technically isolated operating environment. By default, models are used that run on infrastructure located in Germany.
The user interface is also designed to ensure that results can be verified. KAI displays the sources used and enables personal data to be pseudonymised prior to processing.
If we take the crystal ball out of the cupboard and look five years ahead: which administrative service or process should, by then, naturally be AI-supported – and what needs to happen today to ensure we actually get there?
I am certain that, in future, citizens will interact with the administration via digital assistants. Many standardised administrative services will be initiated and completed directly through spoken dialogue.
Within the public administration itself, AI will also transform the way specialist procedures are handled. Staff will carry out many tasks using natural language.
Subsequent requests for additional information should also become significantly less frequent. Checks for completeness and plausibility can already be carried out whilst the application is being submitted.
From AI experimentation to digital infrastructure
The next steps in the use of artificial intelligence in public administration will depend less on whether even more pilot projects are launched. What matters is which applications can be operated on a permanent, secure and scalable basis.
This requires shared technical infrastructure, clear standards, qualified staff and processes in which AI takes on a specific task.
How this transition from experimentation to routine operation can be achieved is also one of the topics at the Smart Country Convention. AKDB has been a partner of SCCON for many years and will be taking part again in 2026.
Visitors will find the company, as usual, right at the entrance to Hall 25 at booth number 400. There, discussions on AI in public administration, secure infrastructures and specific applications can continue directly.
Further information is available at: www.akdb.de/sccon