Participants in the agentic AI panel at the Portfolio Bank Technology 2026 conference agreed there is a sizable gap between financial institutions' ambitions for AI and the practical limits imposed by compliance, security and supervisory requirements. The discussion was moderated by Dojcsák Dániel (Shiwaforce).
Key takeaways
-
Speakers noted that AI can make financial services cheaper while also displacing a substantial share of jobs. Prokop Péter, representing Uniqa Biztosító, said the insurer rebuilt an end‑to‑end claims settlement process with artificial intelligence, has already settled several thousand claims with the solution, and can reduce claims handling costs to one third. He added that 40–45 percent of the required workforce could be replaceable from 2028.
-
Both András (MBH Bank) pointed out that banks operate many different systems and nearly every business area has legitimate ideas for using AI. Those ambitions, however, often diverge from what is feasible under compliance and supervisory constraints.
-
Panelists suggested that cloud service prices may reasonably rise in the future. Strén Gábor (Amazon Web Services) argued that cloud and big tech developments originate from real market and customer needs, so each initiative should start from an actual problem, and solutions moved to the cloud must be continuously modernized.
Technical and regulatory challenges
-
Mádi Gábor (Shiwaforce) emphasized obstacles and risks that complicate AI adoption: past mistakes in public sector digitization where scanned documents were treated as digital records, legacy systems, and the misconception that AI is always the right answer. He also warned that bank secrecy and security impose hard limits, so not every process can be entrusted solely to agents.
-
Kada Zsolt of Magyar Nemzeti Bank (MNB) stressed the need for predictable, stable and robust processes in banking. He noted that AI's non‑deterministic, stochastic nature makes recoverability and rigorous risk assessment particularly important. According to participants, the MNB's supervisory stance is not inherently strict but rather refined and risk‑based.
-
The human factor was repeatedly mentioned as underappreciated: technological development must be accompanied by attention to organizational and workforce issues.
Practical examples and conclusions
-
Uniqa's case demonstrates tangible savings and efficiency gains from AI even alongside existing systems when implemented in an integrated, end‑to‑end manner. Yet many institutions see AI deployments as risky because of legacy infrastructure.
-
The panel concluded that governance for AI applications cannot rely on ready‑made templates: principles and processes should be developed case by case. Technological progress should occur within responsible regulatory frameworks, and cutting‑edge technology alone cannot substitute for strategic planning.
Why this matters
The discussion highlighted that while AI can accelerate digital transformation and reshape cost and risk structures in banking, implementation involves more than technology alone. Regulatory compliance, security, recoverability, legacy system integration and human factors remain critical. Going forward, risk‑based, use case‑specific approaches and careful integration with existing systems will be essential.
Related event
The article also notes the Portfolio AI & Digital Transformation 2026 conference scheduled for November 26, for which registration information was available at the time of the report.
