How AI Is Helping Government Companies in Slovenia Cut Costs and Improve Efficiency

Slovenia's National AI Programme (NpUI) steers public bodies toward practical AI adoption - earmarking EUR 110 million, prioritising reference implementations in public administration, staff training, open public sector data and high-performance computing (including the Vega EuroHPC system) to speed research into live services (Slovenia AI Strategy Report (EU AI Watch)).

Real-world pilots already range from tax‑fraud detection and epidemiological modelling to METIS for early detection of learning problems, while watchdogs and journalists are pushing transparency via the new Public Sector AI Registry to hold institutions accountable (Public Sector AI Registry (EDRi)).

For civil servants and vendors who want to turn these national priorities into day‑to‑day gains - faster decisions, fewer manual steps - practical training like Nucamp's 15‑week AI Essentials for Work course teaches prompts, tools and workplace workflows to make AI adoption safer and more productive (Nucamp AI Essentials for Work syllabus).

Attribute	Information
Description	Gain practical AI skills for any workplace; learn AI tools, prompts and apply AI across business functions.
Length	15 Weeks
Courses included	AI at Work: Foundations; Writing AI Prompts; Job Based Practical AI Skills
Cost (early bird / after)	$3,582 / $3,942
Registration	Register for Nucamp AI Essentials for Work

Slovenia's National Programme for Artificial Intelligence (NpUI) turns strategy into spending - the government has earmarked EUR 110 million through 2025 to move AI from labs into live public services, prioritising reference implementations in public administration, workforce reskilling and a cutting‑edge data and computing backbone (including the Vega EuroHPC system) to accelerate research-to-market pipelines (see the Slovenia AI Strategy Report (EU AI Watch)).

The programme combines short-term pilots (tax‑fraud detection, epidemiological modelling) with longer-term bets on human capital, open data platforms (OPSI) and a National AI Observatory to monitor uptake; Slovenia's deep research roots at the Jožef Stefan Institute and university labs give the programme credible technical muscle and dozens of AI groups ready to partner on deployments.

Policymakers are already refreshing the roadmap - the Ministry of Digital Transformation invited public comment on a 2026–2030 strategic framework - so public bodies and vendors should expect funding and governance to keep evolving as Slovenia builds practical, demonstrable AI that saves time and cuts costs across health, industry and administration.

Attribute	Detail
Public funding	EUR 110 million (to 2025)
Timeline	NpUI 2020–2025; public consultation on 2026–2030 framework
Key priorities	Reference implementations, human capital, HPC, open data, standards
Infrastructure	Vega EuroHPC, OPSI open data platform

Slovenia's approach to shared AI infrastructure is practical and collaborative: the national SLING network and the annual SLING Days supercomputing conference gather researchers, SMEs and industry users to turn raw compute into usable services, with companies such as AformX, Flai and HiSense‑Gorenje sharing real‑world HPC+AI case studies (SLING Days supercomputing conference).

The backbone is the HPC RIVR programme, which placed the EuroHPC Vega supercomputer at IZUM in Maribor to provide a national, open research and industry platform; Vega is explicitly positioned to accelerate AI, HPDA and cross‑border projects (HPC RIVR Vega overview at IZUM).

Policy and OECD summaries underline that this shared model also opens capacity to business - up to about 20% for companies - so public agencies, researchers and SMEs can tap the same high‑end resources.

The result: faster model training, larger synthetic data experiments and mission‑critical simulations on a machine with 6.8 petaflops sustained performance and 240 Nvidia A100 GPUs - a scale that makes previously impossible public‑sector AI pilots suddenly affordable and actionable (VEGA HPC Super Computer (OECD)).

Attribute	VEGA / HPC RIVR
Sustained performance	6.8 petaflops
Peak performance	10.1 petaflops
CPU cores & CPUs	122,800 cores; 1,920 CPUs (AMD Epyc)
GPUs	240 Nvidia A100 cards
Storage & throughput	NVMe Lustre 1PB; Ceph 23PB; 400 GB/s HP storage throughput
Access model	National access for researchers; ~20% capacity available to companies
Host / Location	IZUM datacentre, Maribor (part of EuroHPC)

Slovenia's push for “reference implementations” is the practical hinge between strategy and real savings: the NpUI explicitly targets demonstrable AI deployments across health, Industry 4.0, language & culture services and public administration to move projects from lab prototypes into live services (see the EU AI Watch report on Slovenia's National Programme for AI (NpUI)).

That playbook pairs national coordination with local capacity‑builders - Digital Innovation Hubs and regional alliances that help SMEs and public bodies adopt edge and HPC-backed solutions - a model promoted in EU DIH projects linking hubs, AI vendors and manufacturing firms (CORDIS project page for DIH Alliance for AI-at-the-Edge).

Standards and international cooperation are core too: the programme funds standards work and EU partnerships so reference builds are interoperable and auditable, and civil society scrutiny is rising via tools like the Public Sector AI Registry for monitoring AI use by public institutions, which helps journalists and citizens track deployments.

Concrete pilots - from METIS school‑alerts and tax‑fraud detection to biometric boarding trials that cut processing to seconds - show how reference projects can shrink back‑office time and surface real efficiency gains for government services.

"You do not need AI to see that a pupil has got worse grades in the second semester. You can do that with Excel," said a speaker at the Grounded festival.

Automation and decision‑support are already reshaping Slovenian public administration, but the shift is pragmatic rather than unconstrained: pilots from METIS school‑alerts to tax‑fraud ranking sit alongside tense debates about passenger screening after a system matched some 8,928 flags from roughly 800,000 PNR records, reminding policymakers that speed must meet legal safeguards (see AlgorithmWatch's country review on Slovenia).

Slovenia's Recovery and Resilience Plan locks these lessons into reform - funding modernised e‑services, an Informatics Development Council and large‑scale staff upskilling so civil servants can supervise algorithmic aides instead of being replaced by them - while real delivery platforms like the MDT interoperability backbone and Endava's SI‑CeV secure mail/appointment stack show measurable returns (over €1m saved and 300,000+ deliveries processed) when automation is paired with robust architecture and clear processes.

The practical takeaway for public bodies: automate predictable, non‑discretionary tasks, require impact assessments and human review for final decisions, and use shared, secure infrastructure so decision‑support becomes a tool for faster, fairer public services rather than a source of hidden error.

Article 27 is unique as it proposes an ex ante safeguard where a bill relating to public sector processing of data involving the use of new technologies must include a completed impact assessment on human rights and fundamental freedoms of individuals.

Predictive analytics are already moving from theory to action in Slovenia's health and emergency services, where specialised methods help translate messy records into usable forecasts.

See the IJS Health methods for predictive analytics: IJS Health methods for predictive analytics.

based on inductive machine

The Jožef Stefan Institute supplies the technical muscle - data analytics, predictive analysis, IoT and edge-ready sensing, plus high‑performance computing and decision‑management toolkits - that let hospitals and civil protection teams model short‑term demand and capacity scenarios.

Read the Jožef Stefan Institute AI and predictive analysis overview: Jožef Stefan Institute AI and predictive analysis overview.

Practical examples include epidemic‑forecast prompts that convert model outputs into 14/28‑day hospital demand and scenario recommendations, a vivid operational detail that helps planners move from guesses to concrete staffing and supply choices.

See an example epidemic forecasting prompt for 14/28‑day hospital demand modeling: Epidemic forecasting prompt for 14/28‑day hospital demand modeling.

The payoff for public bodies is clear: earlier, more accurate warnings, faster allocation of scarce resources, and fewer last‑minute crises when surge demand arrives.

Attribute	Detail
Health methods	Inductive machine analysis of clinical databases (IJS)
Research partner	Jožef Stefan Institute - AI, predictive analysis, HPC, sensors
Operational use	Epidemic forecasting converting outputs into 14/28-day hospital demand

Slovenia's National Programme for AI (NpUI) even envisages pilot projects to use AI in cybersecurity and to enhance police efficiency, part of a broader push to detect fraud and cyber threats more proactively (EU AI Watch Slovenia AI Strategy Report); those pilots sit alongside a clear national incident‑management architecture where the Sector for Information and Cybersecurity (SIKV) coordinates responses with SIGOV‑CERT, SI‑CERT, the Ministry of Defence and police, and links into EU networks such as CyCLONe to share alerts and lessons fast (Government Information Security Office SIKV - Information and Cyber Security Division).

Independent reviews also map out how roles and responsibilities are split between agencies so AI tools feed into an organised chain of command rather than siloed watchers (CCDCOE National Cybersecurity Organisation Report - Slovenia).

The practical upshot for public bodies: combining targeted AI pilots with a single, operating centre - down to a physical office at Ulica gledališča BTC 2 in Ljubljana - helps turn anomaly signals into coordinated investigations instead of lost alerts, shrinking the window between breach and response.

Agency	Role
SIKV (Sector for Information & Cybersecurity)	Coordinates national information security and incident management
SIGOV‑CERT	Incident Response Center for government information systems
SI‑CERT	National Cybersecurity Response Center
EU CyCLONe	European Cyber Crisis Liaison Organisation Network - cross-border information exchange

Slovenia's smart‑cities and Industry 4.0 push is already concrete: EU‑backed pilots use AI‑powered digital twins to slice energy costs in maritime and factory settings, with the INNO2MARE project building a live, REST‑API connected twin (hosted at DIGITEH and ISKRA) that runs 48‑hour ensemble forecasts and optimization routines that can hit up to 95% accuracy and balance battery life, grid imports and export opportunities (INNO2MARE marine traffic safety pilot project).

On the factory floor, Digiteh's work combines digital twins with ERP and Lean Production Management to feed real‑time guidance to workers via smartwatches - recalibrating schedules and instructions every five minutes to cut waste and downtime (Digiteh and Siemens AI digital twin real‑time loop overview).

That industrial momentum is mirrored by national scale investment: the University of Ljubljana will host a new supercomputer and the Slovenian Artificial Intelligence Factory to give businesses, researchers and public services a single place to build Edge‑AI and IoT solutions for smart cities and manufacturing (University of Ljubljana announcement: Slovenian Artificial Intelligence Factory), while the NpUI explicitly supports Edge‑AI testbeds such as the Faculty's Factory of the Future demo centre (EU AI Watch report on Slovenia AI strategy and Edge‑AI testbeds).

Partner	Location	Role / Tech
ISKRA	Ljubljanska cesta 24a, 4000 Kranj	Physical systems, real‑time data collection, PV, cogeneration, battery storage
DIGITEH	Tržaška cesta 315, 1000 Ljubljana	Digital twin development, AI models, HMI and integration
University of Ljubljana	Aškerčeva 6, 1000 Ljubljana	Factory of the Future demo lab, Edge‑AI integration, supercomputer / AI Factory host

“AI and digital twin technology, coupled with Plant Simulation, form the backbone of the future of manufacturing, empowering operational efficiency and productivity.” - Dr. Hugo Zupan, Digiteh d.o.o.

Language technologies are becoming a quiet superpower for Slovenian public services: the CLARIN.SI national node, hosted at the Jožef Stefan Institute, preserves and serves over 600 language resources and tools while offering REST services and online concordancers that let archivists, museums and municipal offices extract meaning from mountains of text without reinventing the wheel (CLARIN.SI general information).

The CLASSLA Knowledge Centre stitches together South Slavic corpora, speech datasets and expert helpdesks - its CLASSLA-web corpora already total billions of words and feed Slovenian LLM and ASR work, from benchmarks to the recent LLMs4SSH hub that maps national activity in large language models (CLASSLA: Knowledge Centre for South Slavic languages).

That infrastructure translates into practical savings: faster document understanding, searchable parliamentary archives, and reusable NLP pipelines for translation and cultural-digitisation projects, all backed by a certified, long‑term repository so language data remains usable across projects (CLARIN.SI repository).

Attribute	Detail
Host	Jožef Stefan Institute (CLARIN.SI)
Repository	CLARIN.SI repository (long-term preservation, REST services)
Resources	Over 600 language resources & tools; CLASSLA-web corpora ~11 billion words
Key tools	CLASSLA-Stanza pipeline, NoSketchEngine concordancers, speech corpora
Services	Helpdesk, training (CLASSLA-Express), LLMs4SSH knowledge centre

Building a trustworthy AI‑ready public sector in Slovenia pairs intensive reskilling with clear governance: the National Programme to Promote the Development and Use of Artificial Intelligence (NpAI) explicitly highlights

education and human resources,

calls for updating curricula and promoting lifelong learning, and puts governance, stakeholder engagement and a National AI Observatory at the centre of implementation (Slovenia National AI Programme (NpAI) - DIG Watch resource).

That emphasis sits on a long national tradition - Eurydice notes lifelong education has deep roots in Slovenia and recent reforms target increased adult participation, especially among the less‑educated, so training pipelines can reach every layer of the civil service (Slovenia lifelong learning strategy - Eurydice).

Practical reskilling matters: tools such as RPA, NLP and document‑understanding are already turning stacked paper workflows into automated pipelines, so targeted adult education and clear impact assessments help public servants move from operator to supervisor while governance safeguards build public confidence (AI Essentials for Work bootcamp - Nucamp registration).

Slovenia's smart play on cross‑border collaboration and pooled procurement shows how a small country can buy big tech with shared European buying power: the EuroHPC Joint Undertaking coordinates procurement and funding so national teams don't go it alone (EuroHPC JU selects AI Factories to strengthen Europe's AI leadership), and Slovenia's winning bid will deliver a new AI‑optimised supercomputer and AI Factory run by IZUM, IJS and ARNES that sits near a Maribor hydropower plant and promises roughly 16× the power of Vega - a scale that turns previously unaffordable model training into routine work for researchers, SMEs and public agencies (see the government announcement on the project's selection and scope: Slovenia announces new supercomputer and AI Factory to boost technological power).

For beginners in government or local firms, the practical next step is skills-first: short, workplace-focused training - such as Nucamp AI Essentials for Work 15-week bootcamp - register - makes it possible to translate access to European HPC and AI Factory services into immediate cost savings and safer automation workflows; picture teams running a hospital demand forecast on supercomputer hours purchased through a pooled call instead of buying racks of kit - faster, cheaper, and shared across borders.

Attribute	Detail
EU support / project value	EU contribution ~€67.5M toward a ~€150M project
Hosting / coordination	IZUM (manager) with Jožef Stefan Institute, ARNES and five universities
Operational target	Trial operation expected June 2026
Purpose	AI Factory + supercomputer to serve research, SMEs, public sector