UNIT.City — місце, де люди працюють... КРАЩЕ! Обирай свій простір просто зараз 👉
Олег ОнопрієнкоAI Eng
14 October 2025, 14:06
2025-10-14
The Ministry of Economy implemented an AI module in e-Permit, which accelerated the issuance of licenses
The new module, which uses a large language model, automatically checks all documents for obtaining licenses and facilitates the routine work of ministry officials.
The new module, which uses a large language model, automatically checks all documents for obtaining licenses and facilitates the routine work of ministry officials.
The digital assistant is integrated into the e-Permit system and is designed for preliminary verification of all materials, which was previously performed exclusively manually. The AI module analyzes documents for:
Logicality and integrity of data: checks the correctness of diplomas, the consistency of information, and the relevance of certificates.
Compliance: Checks that education, terms, names, and dates meet the requirements of the license terms.
Data discrepancy: records, for example, discrepancies between the last name on the diploma and the application.
The ministry notes that AI does not make the final decision, but provides the official with structured analytics and specific comments.
How AI works at the Ministry of Economy
AI is currently in pilot mode and concerns a veterinary license, which is complex and multi-component. Previously, checking a single application, which could contain packages of documents for several employees, significantly slowed down the process. Now, an application submitted through "Diya" automatically enters the AI module.
After analysis, the official receives one of two recommendations: “recommend issuance of license” or “reject,” with specific reasons for discrepancies in the report.
Although the pilot launch only concerns veterinary licenses, the system architecture is ready for further scaling to other types of permits and licenses.
Technical stack
The AI module uses a combination of large language models:
Qwen 2.5 7B VL is a multimodal model that extracts text from attached files (OCR) and fixes discrepancies.
Gemma 3 27B is a multimodal model from Google DeepMind that analyzes context, collates data, and generates recommendations.
A feature of the solution is its complete isolation from foreign cloud services. The entire data processing process takes place within the Ukrainian De Novo data center. This ensures digital sovereignty and guarantees that "the state receives world-class power without the risk associated with data being taken abroad."
The development of the AI part was carried out by the MK-Consulting team, coordination by BRDO, and funding was provided by the European Union within the framework of the EU4DigitalUA project. The processing speed of one application is from 2 to 15 minutes.
"We did something we never dreamed of in 2022." How artificial intelligence is changing the state and the army from within — an interview with the creator of the Avengers AI solution that automatically identifies military equipment
Ukrainian answer ChatGPT. How Kyivstar and the Ministry of Digital Economy will build a national LLM for Ukraine: insights and international AI experience VEON