Óbuda University has sought to turn its institutional AI strategy into concrete, measurable programs. According to Eigner György, dean of the John von Neumann Faculty of Informatics and the university’s rectorial appointee for AI, every initiative is paired with specific KPIs: success criteria, deadlines and associated budgets. Early indicators are short-cycle, action-oriented targets — for example organizing a defined-size AI training, launching an internal platform, and collecting and evaluating user feedback.
What ‘AI-nativity’ means: three dimensions
The university views AI-nativity along three axes. First, AI should not be an isolated pilot but integrated into teaching, research and administration. Second, an AI-first mindset asks staff to consider whether a task can be done faster, more reliably or better with AI or automation. Third, AI-readiness means that employees know what tools are available and can use them even if they are not programmers.
Óbuda runs an internal “AI ambassador” system to collect practical use cases, consult with colleagues and monitor what works in practice. Applications that do not achieve expected adoption can be retired, while promising tools — for example an AI-supported career services component — can be adopted and scaled.
Internal GenAI platform: from secure campus use to market interest
A concrete output of the AI Campus program is an internal generative AI platform that provides a single interface for faculty, researchers, students and staff to access multiple models. The system is currently in beta and primarily designed for internal use, but the university has received commercial inquiries and issued price quotations, although no contracts have been signed yet.
The platform is hybrid: some models run on the university’s own servers — important for sensitive institutional data so that data does not leave campus infrastructure — while other high-performance external models can be called when required. The platform acts as a central access and governance layer: users log in with university credentials and administrators can configure who can access which models, with what permission levels and usage quotas.
Potential external customers include other universities, research institutes, state bodies and companies that want a controlled GenAI environment but cannot or do not want to build it from scratch. Eigner emphasized that the aim is not to compete with global foundation-model providers (OpenAI, Anthropic, AWS, Alibaba), but to provide effective, locally deployable access to both local and frontier resources with institutional and data-security controls.
AI in research: where time savings are real
Eigner argues that AI can meaningfully increase research efficiency when a researcher already knows what they are looking for — a hypothesis, a paper, data, or a mathematical model. In such cases AI can speed up literature analysis, extract methodological elements, run simulations or perform validation checks. The Agentic Discovery Platform, developed jointly with the HUN-REN AI Service Center, is a human-in-the-loop research-support system that orchestrates multiple AI agents to help move a research idea toward testable software prototypes.
This is not a conventional chatbot: the system can extract main claims and methods from a technical text, produce an initial runnable demonstration program, perform statistical analysis, and search for data to better understand the problem. Eigner gave a concrete example: building and tuning a computational implementation of a glucose-homeostasis mathematical model previously required days or weeks, whereas with the platform an initial verifiable version and a first draft study outline were produced in one to two hours.
At the same time, AI does not replace complex reasoning, experimental work, field data collection or the human judgment that determines what to research and how to interpret results responsibly.
Work rhythm and mental load
Agent-based systems change not only productivity but work rhythms: an AI agent can work overnight and deliver results in the morning that need assessment and further instructions. This increases throughput but also mental workload, and makes it harder to switch off. Eigner stresses that adaptation is needed not only technologically but also organizationally and educationally.
What AI sovereignty means in practice
Eigner noted that AI sovereignty is often a rhetorical concept. Hungary is not fully sovereign in digital capabilities, and reliance on global services is deep. Rather than focusing solely on building a domestic large language model, realistic national strategy should map technological dependencies and where feasible seek control. The region is unlikely to match the US and China on compute, capital or data scale, but Europe has industrial strengths (for example in specialized manufacturing) that could underpin regional AI capacities.
For Hungary, the pragmatic path is not competing head-on in foundation models but understanding and using these technologies under controlled conditions, protecting local data and preparing the economy, education and research for AI transition. This requires compute resources, training, institutional collaboration and governmental support.
Impact on junior IT roles and education
AI automates and accelerates many tasks traditionally done by junior engineers, so entry-level roles are evolving rather than disappearing. Employers increasingly expect graduates to work with AI tools, think in complex systems and take on problem solving beyond classic junior tasks. University curricula typically react slower than the labor market, so Óbuda University has initiated a professional committee proposal: institutions offering IT education would work with government to define strategic educational priorities, review current programs and consider accreditation changes.
The university pursues two educational approaches. First, AI training is offered across all degree programs. Second, AI should be an integral subject area: students should not only learn use cases, but also how systems work, what their limits are, and the ethical and data-protection issues involved. Deeper technical foundations are to be taught — hardware operation, energy efficiency, model optimization such as quantization — and new roles must be prepared: operators and maintainers of AI systems, security specialists, and professionals who understand agent behavior.
Eigner underlines that core scientific and critical-thinking foundations must remain: as tools change rapidly, students need to understand the principles behind the technology so they can adapt. Universities must shorten the time to practical onboarding and provide specialized course material faster so graduates can contribute in industry sooner.
Conclusion
Óbuda University’s AI Campus program — including the internal GenAI platform, the Agentic Discovery Platform collaboration, and curricular changes — aims to make AI a measurable, operational component of daily academic life rather than a rhetorical slogan. The university received support for this article.
