Find
Answers citizens and staff only from the city hall's official documents, in Romanian and Russian, citing the document and the passage.
Products FIR
Management system for city halls Prototype · first place at GigaHack 2026
The question, the document and the person who answers, on the same thread.
FIR — Find, Inform, Resolve — is the management and work system for city halls. It answers only from the city hall's official documents, gathers every channel into a single inbox and takes each request to a responsible person, with its legal deadline.
- 1Finds in documents
- 2Gathers the channels
- 3Resolves on time
FIR
From information to work done.
Inform — gathers
Web, phone, email, the counter and letters, including handwritten ones, all arrive in the same work inbox.
Resolve
The AI proposes the responsible department, a person confirms, and the task receives its legal deadline.
Where it becomes useful.
Relations with citizens
Answers with the source cited. When information is missing or documents contradict each other, the system says so and sends the person to the right contact page.
Petitions and deadlines
Petitions are tracked against the legal 30-day deadline, extendable by 15, and the person who prepares the reply is not the one who signs it.
Internal coordination
Requests between departments with deadlines and escalation, circulars with read confirmation, meetings whose minutes become tasks.
FIR was developed by the MEGA PROMOTING team for the Chișinău City Hall challenge at Deeptech GigaHack 2026. It is a proposal and a prototype, not a system contracted or currently used by the city hall.
FIR in detail
What you can do with this project.
In a city hall, the same question arrives by many routes — by phone, by email, at the counter, in a letter, on the website — and each time someone looks again for the document, the department and the deadline. FIR puts all three on the same thread: it finds the answer in the official documents, receives the request wherever it appeared and takes it to a responsible person who closes it on time.
The system models the real structure of a city hall: directorates and departments, sector offices, suburban town halls, municipal enterprises and subordinate institutions — schools, kindergartens and the rest. In the prototype built for Chișinău, this means 5 sector offices, 18 suburban town halls, 450 institutions and 183 voice agents.
Answers only from the city hall's documents
The assistant answers citizens and staff in Romanian and Russian, only from the city hall's official documents. Each answer cites the exact document and passage. When information is missing or two documents contradict each other, it says so openly and sends the person to the right contact page, instead of filling in from assumptions.
A single inbox for every channel
Requests from the web, the phone, email, the counter and letters arrive in the same place. Letters, including handwritten ones, are read by optical character recognition.
A voice for every department
Each department has its own voice agent. Alongside them there is a mayor's line and a personal voice assistant.
The AI proposes, a person confirms
For each request, the system proposes the responsible department. The assignment becomes final only after a person confirms it.
Legal deadlines and separate signatures
Tasks carry their legal deadline — for petitions, 30 days, with the option of extending by another 15. The person who prepares the reply is not the one who signs it.
Requests, circulars and meetings
Requests between departments have deadlines and escalation. Circulars are sent with read confirmation. Meeting minutes become tasks, with an owner and a deadline.
Public tracking and a Radar for management
Citizens can follow their case publicly, with no personal data displayed. Management has a calm Radar that shows the state of the work without turning every request into an alarm.
Data and operation
What enters the system. What needs to be checked.
- Personal data is masked before the AI
- Before any call to an AI model, personal data in the request is masked. The model works on the text, not on the person's identity.
- The city hall remains the data controller
- The data and documents belong to the city hall, and the city hall remains their controller. The system works according to its structure and rules.
- The AI proposes, people decide
- Assigning a request, the official reply and the signature remain human decisions. The model prepares proposals; it does not decide in place of the official.
- Hosting in the European Union, with a path to sovereign hosting
- Today the system runs on infrastructure in the European Union. The next step can be sovereign hosting at STISC and our own Grai voice model for Romanian and Russian, subject to validated access and the necessary approvals.
From exploration to implementation
How we prepare a project with FIR.
We choose the circuit for the pilot
We start from one department or one type of request, with its real documents, people and deadlines.
We set the reference documents
The assistant answers only from documents approved by the city hall. Their list is agreed with the people in charge.
We configure the roles
On the city hall's structure we set who assigns, who prepares the reply and who signs.
We decide the hosting
Infrastructure in the European Union for the pilot, or sovereign hosting, depending on the institution's requirements and the approvals needed.
Questions worth clarifying.
Is FIR already used by Chișinău City Hall?
No. FIR was developed by the MEGA PROMOTING team for the Chișinău City Hall challenge at Deeptech GigaHack 2026. It is a proposal and a working prototype, not a signed contract. The figures on this page — sectors, suburban town halls, institutions, voice agents — describe the structure modelled in the prototype.
What happens when the answer is not in the documents?
The assistant says openly that the information is missing or that documents contradict each other, and sends the person to the right contact page. It does not fill in the answer from assumptions.
Does the AI decide who handles a request?
No. The system proposes the responsible department, and a person confirms the assignment. The same principle applies to the reply and the signature.
In which languages does it answer?
In Romanian and Russian.
Where is the data hosted?
Today, in the European Union. Personal data is masked before any call to an AI model, and the city hall remains the data controller. Sovereign hosting at STISC and our own Grai voice model are the possible next step.
Can it be used in a smaller city hall?
The structure is configurable: directorates, sector offices, subordinate town halls, enterprises and institutions. A smaller city hall has fewer levels but the same needs — documents, requests, deadlines. A pilot starts from what the institution has.
Illustrative example
A handwritten letter about the street lighting
A use case, without client data or attributed commercial results.
Initial situation
A resident of a suburban town hall sends a handwritten letter: the lighting on their street is not working. The letter reaches the registry.
How it works
The letter is read by optical character recognition and enters the same inbox as requests from the web, the phone and email. Personal data is masked before the model is called. The system proposes the responsible department, and an official confirms. The task receives the petition's legal deadline; the reply is prepared by one person and signed by another.
The result
The resident follows the case publicly, with no personal data displayed. Management sees the state of the work in the Radar, without following every letter individually.
What is required:The city hall's official documents and its structure, with the roles of assigning, preparing and signing. The scenario is illustrative.
Collaboration possibilities
FIR, in the context of your organisation.
A single place where channels arrive
Calls, email, forms and external systems that write to the same record, with explicit rules about what goes back and who has access to what.
Private companies
We define a pilot around a real process: users, data, integrations, costs and acceptance criteria. Expansion follows after evaluating the result.
State institutions and companies
We establish the requirements for accessibility, hosting, data protection and interoperability. Any connection to AGE or STISC services requires validation of eligibility, access and approvals.
These are adaptation scenarios, not statements about existing contracts or partnerships. The proposed functions are confirmed within the project scope.
Discuss a pilotPart of an ecosystem.
AI voice agents
Kallina
We build agents that answer and place calls, use the information the business already holds, and work with your systems.
PlatformVoice research
Grai
Grai is our research direction for speech synthesis and for models adapted to the languages spoken here.
Research & developmentCivic platform
sesizari.md
sesizari.md organises reports about public space: location, category, description and the relevant material.
Public platformWhat would you want to work better?
Tell us about your process. Together we decide what is worth building, what we can connect, and how we check the result.