Márton Suppan, founder and CEO of Peak, argues that the main obstacles to adopting artificial intelligence are organizational rather than purely technological. While many companies focus on questions such as which model to use, whether to build an agent, or which platform to buy, Suppan emphasizes that understanding a company's internal operations is far more important.
Peak started 13 years ago in fintech, building card programs that enabled other companies to offer financial services under their own brand more quickly and cheaply. Building on that experience, the company developed its AI business: it first used AI to speed up its own internal processes, and when those results were shown to clients the response resembled their earlier fintech experience — rather than applause, potential customers asked how they could get the same capabilities. That demand led to PeakX, an AI-agent platform and implementation practice aimed at enterprise requirements.
Suppan stresses that AI projects are not something you can simply package and hand over. He estimates that an AI implementation can be up to 80 percent organizational transformation and at most 20 percent technology. In many cases the technological component is even smaller, because the tool itself already exists and the challenge is to teach people how to use it and align it with corporate processes. This issue is especially relevant for Hungarian small and medium-sized enterprises: research linked to the Peak AI Start 500 program found that many companies already use some AI tools, but this usage often amounts to an employee asking their personal ChatGPT account to draft an email or summarize a text. That is at least an attempt, but it is not a corporate AI strategy.
Another problem is the lack of foundational digital systems. Suppan notes that fewer companies have CRM systems than those that already use some AI tools. It is difficult to implement AI where a company does not know exactly who its customers are, where cases stand in the process, what data is available, and who does what. Accordingly, Suppan sees education as a key issue for the coming years — training should target not only leaders but also employees, because a good AI strategy is ineffective if staff are AI-illiterate, distrustful, or use tools via uncontrolled personal accounts. Suppan believes the state also has significant responsibility in this area.
The conversation also covered the future of banking. Suppan does not expect bank branches, online banking, or mobile apps to disappear within a few years: the banking market changes slowly, and fintech has not eliminated the traditional banking system despite predictions made a decade ago. Rather, many fintech firms have disappeared and some large players have themselves become banks.
Suppan expects AI’s earliest tangible effects in banking to be on back-office processes rather than user interfaces. Much of a financial institution’s operation is still manual, slow, and fragmented: for example, a mortgage process often stalls not because a client cannot meet requirements, but because internal procedures, requests for missing documents, document handling errors, and outdated systems slow things down. If those under-the-hood processes can be sped up with AI and improved process management, banking can become simpler and faster.
The podcast timestamps for topics discussed are: what Peak does (00:35), traditional banking costs and cheaper financial solutions (08:23), why Peak is not going public (14:03), what Peak AI can do (16:05), the similarity of AI’s current state to fintech’s early phase (17:02), PeakX AI and AI implementation (27:16), AI Start 500 findings on Hungarian SMEs’ digital lag (29:28), and how AI could change banking (35:42). Further details of the conversation are available on the company’s podcasts.
In summary, Suppan’s view is that the technology is largely available today; success will depend on understanding internal operations, transforming processes, and educating employees — factors that will determine whether AI truly makes banking faster and simpler.
