State modernisationNovember 2026 · 7-minute read
An administrative leap, not administrative optimisation
Artificial intelligence, entrepreneurial competition and new forms of cooperation make a different kind of state modernisation possible.
Germany’s Federal Testing Act (Bundeserprobungsgesetz) has been in force since 10 October 2026. It opens up new ways to depart temporarily from certain provisions of administrative law and to test alternative procedures under real-world conditions. Until now, state modernisation has mainly aimed to make existing structures more efficient.
The Federal Court of Audit has documented the shortcomings of this approach for years. The consolidation of federal IT services alone cost 836 million euros without achieving the intended widespread use of shared IT solutions. The Deutschland-Stack and stronger central IT governance are meant to overcome such problems. Shared technical foundations are necessary, but they are not enough for an administrative leap.
What sets an administrative leap apart from conventional modernisation is that it does not merely improve existing processes – it overcomes previous limitations. This changes both what state modernisation can achieve and what it demands.
When old limits no longer apply
I spent the first years of my career in trend and social research. One formative insight from that time concerns the conditions for developmental leaps.
They often occur when technological, economic or institutional constraints that used to be unavoidable fall away. Usually several developments come together: technologies reach a new level of performance, costs fall, infrastructure becomes available and long-held assumptions lose their validity.
This pattern can be seen from the electrification of industry to the smartphone. The decisive progress did not come from new technologies alone, but from reorganising existing processes and business models. That takes a willingness to question basic assumptions, to allow experiments and to spread successful solutions quickly.
For administrative modernisation, this raises different questions: Do specialist procedures have to be harmonised before they can be used digitally? Does every application have to remain a separate case? Do citizens need to know which authority is responsible? Does the state have to develop every digital access point itself?
With artificial intelligence, interoperable infrastructures and new operating models, some of these requirements – so far justified on technical or organisational grounds – are no longer inevitable.
An administrative leap is not about organising yesterday’s administration more efficiently, but about overcoming previous limitations.
Design for Simplexity
To achieve this, digital administration must meet three requirements at once: enable citizens to act, enable AI to work and enable the administration to steer. We call this approach Design for Simplexity: as simple as possible for citizens, while mastering the necessary complexity in the background.
Ideally, the administration would understand what citizens need, recognise their entitlements, obtain existing information itself and complete cases without unnecessary queries. For this, AI needs clearly interpretable data, legal bases, responsibilities and decision rules. The administration, in turn, must ensure that all procedures remain lawful, economical, controllable and reliably adaptable.
These requirements do not always coincide. That is where the real architectural challenge lies. Until now, specialist procedures often had to be standardised to make their digitalisation and maintenance economically viable. AI changes this calculation. It can process different documents, data structures and process variants, integrate existing applications and prepare adjustments.
Another existing limitation lies in the way our processes are designed. Over decades, they were organised so that people could understand, learn and carry out every processing step. This requirement still shapes our idea of standardisation and digitalisation today. As AI increasingly takes over the processing, this constraint loses its significance. The infrastructures we are building today will be used for many years and will meet AI systems whose capabilities will foreseeably go far beyond what is possible today. We should therefore stop designing processes solely around the abilities of human caseworkers and design them around the possibilities of intelligent systems instead. The rules applied, the results and the responsibilities must remain traceable – but not every technical intermediate step has to be something a person can carry out. Otherwise, we turn our current abilities into an artificial ceiling for future solutions.
This shifts the focus of standardisation. Binding rules, data semantics, interfaces and security requirements must be uniform and reliable. Specialist procedures, on the other hand, can remain different where this makes sense legally, organisationally or for users. Even additional complexity in the background can be justified if it means citizens are served considerably better.
At the same time, AI can help identify which procedural variants are necessary, which forms of proof cause unnecessary effort and which differences have merely grown historically.
Simplifying every specialist procedure no longer has to be a precondition for digitalising it. Digitalisation itself can become the basis for targeted simplification.
This avoids making technical modernisation dependent on years of negotiation about responsibilities, procedural variants and legal changes, without permanently preserving superfluous complexity.
Entrepreneurship instead of more contractors
Besides new technologies, an administrative leap requires different economic and organisational models. From large transformation programmes in banking, insurance and other regulated industries, we know how strongly results depend on responsibilities and economic incentives.
The state has long worked with private IT companies, but mostly in a classic contracting logic. A service provider paid by effort sells labour. Under a fixed-price contract, it takes on development risks; an operator with a service level agreement is responsible for the agreed quality of operation and certain economic risks.
An entrepreneur who invests their own capital, develops their own product and is responsible for its commercial success in the market follows a different logic.
They improve their product to win and keep customers. They automate because manual processes increase their costs. They integrate new technologies as soon as these offer economic advantages, and they finance further development from a scalable business model. Their success depends directly on how well their offering performs.
For the state, this can be more economical than developing and permanently financing its own applications. The preconditions are a sufficiently large market, working competition and the ability to switch providers.
The tax system shows how it can work: with ELSTER, the tax administration provides the state infrastructure and a free application. Private companies build their own products on top of it, which many citizens pay for voluntarily. Providers such as WISO Steuer or Taxfix compete on usability, features and support. This has created an independent software market with considerable private revenue, while sovereign responsibility remains with the tax administration.
Similar divisions of labour exist internationally, for example at the British tax authority HMRC, with the Swedish digital identity BankID and in the payment infrastructures of India and Brazil. The state guarantees reliable foundations, while independent providers develop their own applications and keep improving them in competition.
Such a division of labour demands considerable governance capability from the state. Legal compliance, data protection, security, interoperability and the avoidance of new vendor lock-in must be ensured. Here the Federal Court of Audit’s criticism also applies: external service providers are no substitute for the state’s ability to steer. Where in-house expertise is lost, dependencies can even grow.
However, the ability to develop applications in-house is not the same as the ability to control their quality, security and cost-effectiveness.
A publicly governed infrastructure with equal access for independent providers therefore combines the state’s ability to steer with entrepreneurial innovation. The preconditions are binding standards, fair access conditions and a clear allocation of responsibilities and risks.
Test, don’t decree
The financial sector has already been through many comparable transformations. What is new is the combination of today’s AI capabilities, shared digital infrastructures and independent entrepreneurial business models.
The frequently criticised unreliability of AI also exposes a problem in how we have understood decision certainty so far. When AI handles the same task several times and comes to different results, that causes unease. But what would have happened if, in a tender, we had hired the second-placed service provider, used a different project team or carried out the same analysis several times? Different, sometimes contradictory results would have been anything but unusual.
We have grown used to this uncertainty without measuring it systematically. With AI, it becomes visible and verifiable. What matters, therefore, is not repeatability alone, but the demonstrable quality of the results. Where the law requires unambiguous decisions, this reliability must be ensured technically. In other areas, we can learn to deal with uncertainty more consciously and more effectively.
No one today can reliably determine which architecture and which operating model will perform best in the long term under these conditions. In particular, there is a lack of solid experience of how far AI actually reduces the costs of integrating, adapting and maintaining complex specialist procedures. That makes experiments under real-world conditions all the more important.
The federal government and the states now have various legal instruments for testing. They make it possible to try alternative approaches without first having to resolve every organisational and legal question conclusively.
Different models can be compared directly: prior harmonisation versus AI-supported integration, state-developed applications versus independent market products, conventional procurement versus self-managed operation.
What matters are measurable results: time to productive use, cost per completed case, degree of automation, error rates, user satisfaction and the effort needed to adapt to legal changes.
This will probably show that different tasks also need different solutions. Standardised high-volume procedures make different demands than rare, complex discretionary decisions. Some services lend themselves to market products, others to individual applications or jointly operated public infrastructure.
The shared architecture must make this diversity possible instead of ruling it out with a single solution fixed in advance.
Developmental leaps rarely follow a complete master plan. They need room for different approaches, short learning cycles and the ability to spread successful solutions quickly.
From administrative procedure to actual delivery
How great the need for action is can be seen in our study, prepared together with eastern German business associations, Ostdeutschland – Deutschlands Chance für Europas Aufbruch (Eastern Germany: Germany’s opportunity for a new start in Europe).
Of the 138 entrepreneurs and executives surveyed, 98 per cent believed that faster authorities, digital procedures and shorter processing times are not enough on their own. Often the underlying rules themselves prevent economically sensible projects from being carried out. 82 per cent were in favour of regulatory testing grounds limited in time, place and scope.
Easier to build, easier to invest, easier to employ: what matters is not just how quickly applications are processed, but also the conditions under which projects can be carried out at all.
Given high public spending and tax burdens, demographic change and weak economic growth, productivity gains in public administration have long ceased to be a purely organisational matter. They affect the country’s economic performance and the state’s capacity to act.
An administrative leap creates the opportunity not just to manage previous limitations more efficiently, but to overcome them. Practice will show which approaches succeed.