The Ministry of Education: Solving Specific Problems for an AI-Native Team

Public institutions handle large volumes of documents, records, and some teams end up working with that volume more directly than others. Our most interesting work with the Ministry of Education, Research, Development and Youth of the Slovak Republic was with their analytical unit. This was a team that didn’t need an introduction to AI. They already had some experience, so instead of a basic course, we built training and consultations around their own, specific problems.

Their analytical unit, the Education Policy Institute, carries out the ministry’s analytical and strategic work: producing analyses and forecasts, and feeding into how education policy gets shaped. Their approach starts from data, drawing on national and international sources and bringing good practices from abroad into concrete recommendations, while coordinating with organizations under the ministry and with other ministries. That work already involves programming, language models, and machine learning, so what they needed from us wasn’t just a basic introduction, but a way to deepen what they already knew.

We ran a series of trainings for the unit covering machine learning from the ground up, working with large documents using LLMs, generative AI and neural networks in practice, and predictive methods for time series.

From there, we moved into individual consultations built entirely around their own work, including specific use cases like analyzing large volumes of text data from questionnaires. We talked through where AI could realistically fit into what they do, worked out a proposed approach for clustering sentence embeddings with k-means, discussed using Retrieval-Augmented Generation, looked at local, on-premise tools for handling sensitive documents with LLMs, and covered the regulatory questions tied to an experiment they were planning with the OECD.

Alongside the work with the analytical unit, we ran AI literacy training for staff across the ministry, along with workshops on personal data protection and AI regulation. Based on how a sample of employees actually worked, gathered over three in-person meetings, we built two tailored trainings on prompting and on searching, extracting, and analyzing information with AI. Ministry staff also joined several external expert sessions on the AI Act, prompting, and neural networks, and we closed the collaboration with an assessment of the ministry’s overall readiness to use AI.

The ministry got what most institutions actually need, literacy for people just starting out, and real technical depth for the team that was already using AI in its work.