From Weeks to Minutes: Valentyn Shyrobokov on AI's Impact on Automation

Transitioning to modern platforms has always been a significant challenge for companies operating on legacy systems. For many financial institutions, this barrier not only slows down their operational efficiency but also hinders their ability to adopt modern, AI-driven technologies. According to Statista, the financial industry invested an estimated $35 billion in AI in 2023, with $21 billion attributed to banking. This growing reliance on AI underscores the importance of solutions that not only automate workflows but also facilitate seamless migration for new clients, enabling them to leverage innovative tools without disrupting existing operations.

Valentyn Shyrobokov, a Senior Java Developer at SAPIENS, is at the forefront of solving these challenges. Over the past five years, he has contributed to the development of solutions such as Model.AI and Integrate.AI for the Sapiens Decision product. The Sapiens Decision Integrate.AI platform enables business users to integrate machine learning models into decision-making models.

Combined with Sapiens Decision’s Model.AI, which uses generative artificial intelligence to automatically transform natural language into decision-making models, enterprises can accelerate their adoption of decision automation and achieve stronger business results. Valentin’s expertise also includes creating a specialized File Appender for logging in legacy systems, as well as a request dispatch library called Action Dispatcher, which serves as a transitional bridge during system modernization. His significant contributions have been recognized by his membership in IEEE and IAHD, further highlighting his role in advancing industry standards.

  — Valentin, today, automation and AI are transforming nearly every industry. How would you, as an expert with years of experience, assess the current state and future prospects of these technologies? — We are at a pivotal stage in the evolution of automation. Previously, the focus was primarily on optimizing routine operations, but now we see the emergence of systems capable of making complex decisions under uncertain conditions.

This is particularly evident in the financial sector, where the requirements for accuracy and decision-making speed are exceptionally high. That is why our solutions, such as Model.AI, Integrate.AI and ALE (Automated Logic Extraction), were developed with the potential for application across various industries where the precision of automated decision-making processes is critically important.  

During your time at Sapiens, you created Model.AI, which has significantly strengthened the company's position in the advanced technology market. Do you have specific examples of Model.AI implementation? What results did your clients achieve?  — The most illustrative example was the implementation of our solution at a major American bank, ranked among the top ten in the country.

After Model.AI was integrated, the time required to create a new decision model was reduced by 90%. It used to take weeks of manual labor, now – just minutes. Another example is an insurance company that, thanks to the Model.AI subsystem, was able to deploy Sapiens Decision. By leveraging Sapiens Decision, they cut the involvement of IT specialists in launching a new financial product by 90%, significantly accelerated time-to-market, and increased customer efficiency by 70–90%. Notably, the transition was carried out gradually, without disrupting ongoing operations — a key requirement of the client.

  — You mentioned ALE (Automated Logic Extraction) as one of the key developments. This solution has helped expand Sapiens’ client base by enabling more companies to modernize their systems efficiently. Could you explain why automatic logic extraction is so important and what makes your solution unique? — ALE addresses one of the most challenging tasks in modernizing enterprise systems — the extraction of business logic from legacy code. The uniqueness of our approach lies in combining static analysis with machine learning methods to uncover implicit dependencies within the code. This enables us not only to replicate existing logic but also to optimize it during migration to a new platform. As a result, clients gain not just a modernized system but also enhanced performance without risking the loss of critical business logic.

  — You are also the author of unique developments such as File Appender for logging and Action Dispatcher, which have had a significant impact and spread throughout the industry. Could you explain the key problems they solve and why these solutions are considered groundbreaking? — Often, client solutions are based on costly licensed systems whose functionality is only guaranteed when using specific versions of Java and certain virtual machines. This significantly limits the range of available tools. In such cases, specialized solutions need to be developed to meet strict customer requirements. The uniqueness of the LimitedDailyRollingFileAppender lies in its ability to update the main log file at a user-defined frequency while also limiting the maximum number of files. This approach minimizes file operations and CPU usage, and ensures compatibility with older versions of Java.

The innovation of the Action Dispatcher library lies in its declarative approach—via annotations—to defining HTTP request routing, performing parameter conversion and validation, and handling exceptions. In addition, the library supports request logging at every stage of processing. This solution is both lightweight and highly efficient, which is particularly important for systems handling millions of transactions. Such an approach has already delivered an unprecedented level of performance in several major projects.

  — In the context of migrating from legacy systems, how do your solutions ensure security and continuity during the process? — Security and continuity are the cornerstones of our approach. We have developed a specialized, step-by-step migration methodology in which each phase undergoes multi-level testing. For instance, our testing module can generate hundreds of thousands of tests covering all possible scenarios of the decision-making model.

These test cases can be exported and imported into the system, and running them allows us to compare actual outcomes with expected results—ensuring that any discrepancies are detected before they affect the client’s operations. Meanwhile, ALE and Model.AI enable the generation of high-quality decision-making models based on the client’s documentation and code, thus providing a seamless transition to Sapiens Decision.

  — You are an active partner of prestigious international associations, which not every IT specialist can get into without demonstrating high achievements in this field. How does your participation in professional organizations such as IEEE and IAHD influence the development of your solutions? — It provides us with a unique advantage of early access to the latest research and practices in AI and fintech. For example, the architecture of Model.AI is largely based on recent studies in federated learning presented at IEEE conferences. Participation in IAHD allows us to incorporate the expertise of specialists from related fields, which is especially important when working with complex financial systems.

  — How do you assess the prospects for AI development in the financial sector in the coming years? — We are moving toward the creation of self-adaptive systems capable not only of automating existing processes but also of optimizing them in real time. Over the next 5–7 years, we expect the emergence of platforms capable of predictively adjusting business processes based on market trend analysis. Ensuring the accessibility of such technologies for organizations of all sizes will be critically important — this will create a more competitive environment and accelerate the digital transformation of the entire industry.