Robotic Process Automation is yesterday’s news; generative and agentic AI are changing the game today regarding streamlining workflows. The reason is simple: while RPA is broadly limited to automating repetitive, rule-based tasks, the arrival of large language models has opened the door to a dimension shift.
GenAI can analyze complex patterns and generate original content, whereas agentic AI enables systems to make decisions, learn, and adapt autonomously.
“This is the dawn of a whole new era,” describes Máté Jendrolovics, founder and CEO of Intuitech, on the significance of the technology.
As he explains to the Budapest Business Journal, up till now, only those processes that could be grasped in line with the rules of mathematics could be automated. Most workflows are not like that, but the emergence of Gen AI and agentic AI has shifted the goalposts.
Building an agent to perform a specific task accurately requires a complex toolset and know-how. But apart from completing certain actions, like collecting data from pdf files or sorting and validating inbound communication and documents, they need to be able to speak to other agents as well, Jendrolovics notes.
“We have learned how to build such efficient agents fast within a given domain. Then we pick a workflow like complaints or claim management, and we deliver it as end-to-end solutions to our customers,” he says.
What matters even more is that they have learned to put together agents that comply with banking and insurance standards, a fact that “is just so far from being self-evident,” the expert points out. It matters because vendors face massive entry barriers in the industry.
Avoiding ‘Innovation Theater’
“Without accuracy, reliability and transparency, the whole thing is just innovation theater, not a solution. If all those factors are in order, only then can you talk about business impact,” the CEO notes.
Jendrolovics recalls that digitalization’s primary initial purpose was to have manual tasks performed digitally, which was followed by the ambition to make it all more convenient, hence mobile compatible. The next big leap is automating end-to-end workflows, which is what Intuitech is taking on.
Given the complexity of the job, relevant expertise is scarce. Intuitech leads the way in the field by dint of having multiple agents in production with proven results, while most players are still working on their proof of concept and pilot programs.
Jendrolovics has been building digital banking solutions in Hungary and abroad for the past six years and is capitalizing on that knowledge to offer solutions to players in the financial industry that really make an impact.
“To deploy our solutions on a large scale within an organization, we need to provide transparency; it can’t be a black box,” he says. That is an oft-heard criticism of AI systems where no one can tell you why AI does what it does.
“We, therefore, build software that checks that [transparency] box,” the CEO adds, highlighting the necessity of having such an approach. Indeed, Intuitech’s agents don’t just perform tasks autonomously; upon request, they will also explain all the decisions they have made with detailed reasoning.
That provides an opportunity for “human in the loop” experts to validate and even change the output of the virtual agents. The implied result of this progress is that loan officers or complaint managers should get ready to assume a new role overseeing “digital workers” that can be taught concrete tasks and can effectively be treated as trainees.
Working the Workflow
It helps that the main structure of workflows is similar across the whole financial industry; once you have that covered, customer-specific needs can be handled more easily. Such as, for instance, training an agent on an insurance company’s operations manuals, product catalog, and internal processes.
This reusability factor also boosts scalability and flexibility. The latter is key as core banking systems are typically complex to change. The agentic AI approach offers a way around that obstacle.
“We need to build agents that we can deliver within 4-8 weeks each for financial institutes,” Jendrolovics says. To ensure that, their agents are built on top of legacy systems so that they can be implemented with little or no changes required from their clients.
Intuitech’s CEO notes that providing accurate operations and creating real value is the most challenging part. “You need deep domain expertise for that,” he says. There are two types of problem-solving: you either target customer service across industries or just deal with banks. It doesn’t go both ways.
“The primary tech issue for banks is accuracy, but if you can handle that, which is the case with Intuitech, the sky’s the limit.” In fact, the firm is cooperating with a Stanford spinoff that writes code that tries to manipulate Intuitech’s AI agent into providing false information just to test the safety limits of their solution.
The opportunity is underscored by a McKinsey report that estimates the value creation of GenAI in the global banking sector at USD 200 billion-340 bln, the equivalent of 9-15% of banks’ profits.
“What inspires me is to elevate complex banking or insurance processes and critical business functions to a new dimension because that’s how you create the most value,” Jendrolovics concludes.
This article was first published in the Budapest Business Journal print issue of March 7, 2025.



