Citizen reports, a virtual assistant and maps for your city hall.
We build the citizen report platform with map and dispatch, the virtual assistant that answers from the city hall's documents, public GIS maps, the website and the mobile app. The AI runs on enterprise models with EU data processing or on open-source models installed on the city hall's own server, so there are no per-user licenses.
Where a city hall loses time
The problems are the same in villages, towns and cities. Only the volume differs.
Reports arrive on every channel
Phone, email, Facebook, the registry office. There is no single list, deadlines are hard to track, and citizens don't know whether anyone handled their issue.
The same questions, every day
Opening hours, taxes, required documents, forms. The answer is already on the website but hard to find, so people call or come to the counter.
GIS data stays in the office
Many city halls have their zoning plan or utility networks in GIS format, yet citizens and companies still go to the counter to learn what can be built on a plot.
Expensive subscriptions per module
Separate annual licenses for each function, costs that grow with every update, and data that is hard to get back from the vendor.
What we can build for a city hall
Four components that also work on their own. You can start with one and add the rest once the first is in daily use.
Citizen report map and dispatch
Citizens pin the issue on a map, add a photo and get a registration number. In the dispatch panel, staff see every report, route it to a department, reply from templates and track deadlines. Citizens see the status without calling.
AI proposes the category, the department and a public text without personal data. Staff confirm.
City hall virtual assistant
Answers citizens from documents the city hall has published: council decisions, regulations, taxes, opening hours, required documents. Shows the source of every answer. When it can't find the information, it says so and points to a department. An internal version for staff covers procedures and internal documents.
Semantic search (RAG, vector database) over the city hall's documents only, not the internet.
Public GIS maps
We publish the GIS data the city hall already has on a web map: zoning plan, utility networks, public property, green spaces. Citizens search an address, see the zoning and download an information sheet, which does not replace the planning certificate.
We start from existing QGIS projects, GeoPackage or Shapefile data, in Stereo 70.
Website and mobile app
The city hall website rebuilt to be accessible, fast on phones and easy to navigate. A mobile app for iOS and Android for citizen reports and notifications.
WCAG 2.1 AA accessibility. The mobile and web apps share the same codebase.
Enterprise or open-source AI: by budget and by data
The AI components (the assistant, report triage) can run on two kinds of models. We choose based on actual request volume and on how sensitive the data is.
Enterprise models
Claude, GPT, Gemini, with EU data processing
- Best answer quality in Romanian
- Pay for actual usage, with a monthly cap that stops spending automatically
- The account can be in the city hall's name, billed directly
- Submitted data is not used to train the models
Open-source models
Gemma, Llama, Mistral, on the city hall's server or in an EU data center
- Data never leaves the infrastructure the city hall chooses
- No per-question cost, only the server
- Well suited to background tasks: report triage, hiding personal data, summaries
- Real-time conversations with many citizens need a GPU. We tell you up front whether it's worth it
Safety does not depend on the model. Personal data is filtered in code, model outputs go through automated checks, and every AI proposal in dispatch is confirmed by a staff member.
How we work with a city hall
Order matters: first we clarify the procurement, then we build.
- 1
1. Call and demo (60 min)
You see the platform on sample data and we agree which problem to solve first. At the end you know whether it makes sense to continue.
- 2
2. Procurement framing, in writing
With the procurement department we clarify the CPV code, the value and the procedure, usually a direct award. We don't start work until this is clear in writing.
- 3
3. Setup on the city hall's data
We load the documents, the GIS layers and the report categories. The platform runs on a city hall subdomain, for example map.your-city-hall.ro, with data hosted in the EU.
- 4
4. Training and pilot
We train the staff who work in dispatch and those who update the documents. We start with a pilot in one department, then expand.
- 5
5. Operation and monthly support
Updates, backups, monitoring, technical support and refreshing the documents the assistant answers from.
Data, GDPR and the AI Act
A public institution has more obligations than a company. We handle them in the design, not at the end.
Data stays in the EU
Hosting and AI processing in EU regions. The city hall remains the data controller, and we sign the processing agreement required by Article 28 GDPR.
Citizens know they are talking to an AI
The assistant introduces itself as AI from the first message, as Article 50 of the AI Act requires from 2 August 2026.
Personal data stays off the public map
Names, phone numbers or ID numbers in a report's text are detected and hidden before publication, and staff approve the public text.
The data belongs to the city hall
On request or at the end of the contract, the city hall receives its data in open formats (CSV, GeoPackage), with no export fees.
Frequently asked questions
Can we start with a single component?+
Yes, and we recommend it. It usually makes sense to start with citizen reports or the virtual assistant. The rest is added once the first component is in daily use.
We already have GIS or a digital zoning plan from another vendor. Can it be used?+
Yes. We start from the existing files (QGIS projects, GeoPackage, Shapefile) and don't replace what works. The public map and citizen reports are added on top of the current data.
How much does it cost?+
It depends on the components and the size of the city hall. The quote is in lei, after a call, and fits a direct award. Usage-based AI cost is separate and has a monthly cap.
Can the AI make mistakes?+
Yes, any model can. That is why the assistant answers only from the city hall's documents, shows the source and says when it doesn't know. In dispatch the AI only proposes the category, department and public text; the decision stays with staff.
Where does the platform run?+
The standard version runs in the cloud, with EU data, on the city hall's subdomain. If the city hall has its own server and prefers an on-premise install, we deploy there, including open-source AI models.
What experience does NeuroAI have with city halls?+
We are a young company. That is why you see the demo before any discussion about money, and we propose a small pilot with signed acceptance before a larger commitment.
Related services
We build the same components for companies too.
Want to see the demo?
In 60 minutes we show you the platform on sample data and agree together which problem it should solve first in your city hall.
Book a call