The Cultural Funds Platforms operated by the Free and Hanseatic City of Hamburg represent a notable deployment of artificial intelligence in public administration, applying a hybrid pipeline of classical machine learning and generative AI to accelerate the processing of financial aid applications for the cultural sector. The system was developed by Hamburg's finance ministry (Finanzbehörde Hamburg), specifically its digital innovation lab known as Kasse.Hamburg, to administer two major federal funding programmes: the Sonderfonds des Bundes fuer Kulturveranstaltungen (Special Fund for Cultural Events), a EUR 2.5 billion programme created in response to the COVID-19 pandemic, and the Kulturfonds Energie des Bundes (Federal Cultural Fund for Energy), a EUR 1 billion programme established to support cultural institutions facing sharply rising energy costs. Hamburg was selected by the German federal government to lead the technical implementation because of the city's modern IT landscape and digital capabilities, and the resulting platform serves all 16 German federal states through a single unified technology infrastructure.
The AI system is built on SAP Business Technology Platform (SAP BTP) and leverages SAP AI Services to perform three core functions within the aid-application processing workflow. First, it provides automated document classification and evaluation, using the SAP Document Information Extraction service to process the large volumes of supporting documentation submitted by applicants, including proof of event costs, revenue statements, energy consumption records, and identity documents. The system classifies incoming documents by type and extracts structured data fields from unstructured PDF and scanned-image attachments. Second, the SAP Business Entity Recognition service performs automated extraction of key entities such as names, addresses, tax identification numbers, and bank details from application materials, enabling data reconciliation and validation through plausibility checks that reduce errors and help detect potential fraud. Third, the platform incorporates generative AI capabilities through the SAP AI Core generative AI hub, which drafts text templates and correspondence based on the specifics of each application. These draft communications are then reviewed, edited, and approved by human caseworkers before being sent to applicants, ensuring that no AI-generated content reaches applicants without human validation.
The technical architecture follows a hybrid approach, combining classical ML models for document classification and structured data extraction with foundation model capabilities for generative text drafting. The platform was developed as a side-by-side extension on SAP BTP, which enabled rapid deployment: the initial cultural events fund platform was built in approximately three weeks, and the subsequent cultural energy fund platform was launched in just two weeks by building on the existing infrastructure. The solution integrates with Hamburg's core finance software, SAP S/4HANA, for processing aid payments, and uses SAP HANA Cloud and SAP Analytics Cloud for map-based data visualisation and real-time reporting. The German federal tax office authentication service, ELSTER, was integrated via the API Management capability within SAP Integration Suite to help guard against fraud.
The platform operates under a human-on-the-loop (HOTL) oversight model, where caseworkers retain authority to review, edit, and approve all AI outputs before any administrative action is taken. This is particularly important given that the AI system processes personal and sensitive data, including financial details and identity information, all subject to GDPR and the German Federal Data Protection Act (BDSG). The system's generative AI hub capability was specifically designed to avoid sharing sensitive applicant data with third-party large language model providers, keeping data processing within the SAP infrastructure hosted in an EU cloud region to maintain compliance with EU data sovereignty and GDPR localisation requirements.
The platform was developed through a partnership model coordinated by SAP's Preferred Success offering, which provided architecture design expertise and project management. Dataport, the public IT service provider for several German federal states, was responsible for back-end integration, running the solution, and providing technical support. D-LABS GmbH contributed the user experience design for the applicant-facing portal. This collaboration between Hamburg's finance ministry, the federal government, SAP, Dataport, and D-LABS enabled the rapid deployment and scaling of the platform.
The outcomes have been substantial. Approximately 7,000 cultural creators submitted nearly 100,000 applications through the cultural energy fund platform alone, and a total of six million documents were automatically evaluated and classified across both programmes. The AI and generative AI capabilities considerably reduced the person-hours required for application processing by providing automated assistance with data access, request review, suggestion generation, and text block editing. This was particularly valuable for processing energy fund applications, where many caseworkers lacked specialised knowledge of the energy market; the AI capabilities helped them more quickly assess energy consumption data presented in applications. The real-time reporting gave state authorities and the federal government important information on requested budget amounts and application status. According to Budget Director Arne Schneider of the City of Hamburg, the AI capabilities in SAP BTP and SAP AI Services allowed the authorities to process cultural fund applications far more quickly than would otherwise have been possible, and AI-based decision support made the volume of applications manageable by a limited workforce. The project has been recognised as a best-practice model for rapidly deploying AI-enabled aid distribution platforms that could be adapted for other sectors or crisis situations in the future.