In many municipalities, digitalization currently means more online forms, but hardly any relief in daily operations. With the amendment of the Online Access Act (OZG) and the guidelines of the Single Digital Gateway Regulation (SDG), the focus is shifting to the scalability of services. It is no longer enough to only provide applications as static online forms; the expectation is pointing towards dialog-based interactions without major barriers.
In many municipalities, the chatbot is evolving from a pure information channel into an integral element of digital citizen service. It combines information, preliminary checks, and structured routing of requests. Crucial for success are not the linguistic skills of generic AI models, but process reliability and data sovereignty.
Reduction of bureaucracy presents municipalities with new digital requirements
In addition, with the Fourth Bureaucracy Relief Act (BEG IV), which has largely been in force since January 1, 2025, the reduction of bureaucracy is becoming very concrete for the first time – also for cities and municipalities. The elimination of written form requirements, the stronger use of text form, and reduced documentation and proof obligations are noticeably changing expectations of municipal administrative processes. For administrations, this means: Information must be made digitally understandable, legally secure, and accessible with low barriers. The challenge lies less in the legal situation itself, but in the practical implementation on site – especially where citizens expect guidance, explanation, and reliability. Digital chat solutions are thus increasingly moving into focus as a structural element of modern administrative work. This is where modern chatbot solutions come in as a structural aid in the initial citizen contact.
From information to process support: The evolution of citizen dialogue
Previous chatbot solutions in municipalities often served as purely informative signposts. They answered questions about responsibilities or linked to PDF documents. These signpost chatbots (or simple FAQ chatbots) save time on the website, but not in the specialized departments. Citizens are sent from subpage to subpage without their request actually being processed. The current OZG development demands deeper integration to enable seamless processes free of media breaks.
Supporting OZG levels of maturity: From information to transaction
A chatbot systematically integrated into administrative processes helps achieve higher levels of maturity in the OZG. The focus here is on the efficient preparation of processes - not their final assessment:
The bot delivers precise answers based on your administrative data, rather than leaving citizens alone with confusing lists of links (information is only provided).
Requests are pre-qualified and data is validated in such a way that they can be seamlessly transferred to internal departments (e.g., in civil registration or business registrations) (transactions/applications are prepared).
It is about sensible support instead of substitution: Relieving repetitive standard inquiries serves to secure capacities for personal and complex individual case advice. Legal review, discretionary decisions, and special cases remain with humans or caseworkers through editorial dialogues.
Control instead of coincidence: Why AI in citizen service must be limited (RAG and hybrid architectures)
Generic AI systems that are based solely on public web knowledge are unsuitable for official information. The danger of "hallucinations" – i.e., factually incorrect but plausible-sounding answers – is unacceptable for legally binding information.
Controlled knowledge base instead of free text generation with Mercury Intelligence
Instead of using uncontrolled web scraping, Mercury.ai uses among other things an orchestrated process (Retrieval Augmented Generation - RAG):
|
|
|
Only explicitly approved documents, bylaws, and FAQ lists serve as sources. Simply put: The chatbot is only allowed to say what was approved beforehand. Nothing else. | Every response from the AI chatbot is internally traced back to the underlying document version. This supports the traceability of administrative actions. | Where processes are legally strictly defined (e.g., application tracks), deterministic dialogue guidance takes effect. The generative AI handles language understanding, but the process logic is still not left to chance: firmly defined and traceable. |
Infrastructure integration: XZuFi and the reality of the federal structure
Crucial for implementation in municipalities is the support of German administrative standards such as XZuFi and FIM. The chatbots and the Conversational-AI-Platform of Mercury.ai are designed to map these standards. Many solutions are not designed for these specific administrative standards because they have little experience with municipalities, cities, and communities.
Synchronization with editorial systems
A chatbot in public administration must be integrated into the existing data landscape:
Interface connection: Mercury.ai enables the import of service descriptions from the Content Management System (CMS). As soon as changes are approved in the editorial systems of the states or municipalities, they can be imported into the chatbot's knowledge base in defined cycles.
Regional logic: By using the Official Municipality Key (AGS) and stored responsibility logics, the chatbot can direct citizens specifically to the locally responsible agencies – also taking county structures or special-purpose associations into account.
Accessibility and inclusion: BITV 2.0 as a foundation
Public software must be accessible to all citizens. Compliance with BITV 2.0 (based on EN 301 549) is therefore not an optional feature, but a core criterion.
Accessible design and assistive functions
The chatbot widget is optimized for screen readers, fully keyboard-operable, and meets the contrast specifications of WCAG 2.1.
The chatbot enables communication in numerous languages, which facilitates access to information for many citizens.
Mercury.ai offers assistive functions to transform complex texts into drafts of Plain Language (Leichte Sprache). Important: Since Plain Language is standardized (e.g., DIN SPEC 33429), these AI drafts serve as a valuable working basis for the obligatory human review and quality assurance.
Data sovereignty and legal security
In addition to GDPR compliance, avoiding a one-sided technology lock-in is critical to success for German authorities. Mercury.ai consistently follows, among others, the strategy of the IT Planning Council to strengthen Digital Sovereignty: The dependency on individual IT service providers must be reduced to ensure the long-term capacity of the administration to act.
Security "Made in Europe"
|
Differentiation from international models
While international AI models like those from OpenAI (ChatGPT) are linguistically impressive, they are not designed for the specific requirements of German administrations. For global language models, specific German administrative standards are simply irrelevant. They are not trained on German administrative standards such as XZuFi, FIM, or OZG process logic and cannot natively map their structural requirements.
The critical difference: While generic AI only generates text based on probabilities, Mercury.ai ensures legally secure and standard-compliant delivery of information that corresponds one-to-one with municipal workflows.
Conclusion: Digital transformation with a sense of proportion
A chatbot for your administration is a tool to strengthen the operational capacity of agencies in times of skilled labor shortages. However, it is not a "plug-and-play" product, but requires neat integration into administrative processes and data standards. Mercury.ai is your partner to unite innovation with the necessary administrative discipline: Sovereign in hosting, controlled in AI generation, and inclusive in design.
Next step: Together, we will identify the processes in your administration with the highest potential for automation. Schedule a non-binding consultation.
