Speed up approval processes without relinquishing responsibility

Challenge: Complex Procedures and Limited Capacity

Planning and permitting procedures are among the most time-consuming processes in the German public administration. Extensive application documents must be thoroughly reviewed, checked for plausibility, and evaluated from a legal perspective. Supplementary submissions are coordinated, and the basis for decisions is prepared. In many government agencies, all of this is still done largely by hand, tying up resources that are needed elsewhere.
The problem is rarely a lack of expertise. It is the sheer volume of documents and review steps that slows down every process before a qualified decision can even be made.

Solution: AI as a review assistant, not as a decision-maker

As a technology partner of PwC in the SPARK project, VAGO has contributed to the development of AI-supported assistance modules that provide structured support for the review process in approval procedures. The system automatically analyzes application documents, extracts relevant content, checks for formal completeness and plausibility, and provides case workers with prepared review suggestions.
The AI does not make administrative decisions. It processes information in a way that allows humans to review it more quickly and thoroughly. All AI suggestions can be accepted, supplemented, or rejected entirely. The responsibility for decision-making remains with humans at all times.
The system is supplemented by AI-supported legal doctrine, which is linked to legal databases and prepares legal requirements in a structured format for technical review. In the future, additional modules for substantive review, legal assessment, and decision drafting are planned.

Technical Highlights: Open Source, Reusable, Digitally Sovereign

The SPARK modules are designed as an agent-based assistance solution and are made available as open source on the openCode platform. They are freely available for reuse by government agencies, the private sector, and civil society. Development took place as part of an innovation partnership and was funded by the Climate and Transformation Fund. The system is currently in beta; the first release includes the modules for content extraction, formal completeness checks, and plausibility checks.

• Agent-Based Analysis and Audit Preparation
• AI-supported legal doctrine with integration into a legal database
• Open-source release on openCode

Result: Less routine work. Faster procedures. Clear accountability.

Approval authorities that use SPARK significantly reduce the manual effort involved in the initial review and structuring of extensive application documents. Case officers receive a clear overview of the process more quickly, along with notifications of missing or contradictory information and prepared review proposals. SPARK thus makes a concrete contribution to accelerating planning and permitting processes, where transparency and human oversight are non-negotiable.