AI Agents Instead of Programmers: New Opportunities for Banks

Developers demonstrated the product’s potential at the 13th International Forum “All About Banking Automation,” and customers have already had a chance to evaluate it.

Agentic Development

Artsofte, a Skolkovo resident, has built a system in which an intelligent algorithm takes on entire stages of banking application development, rather than just writing individual snippets of code. It designs data models, builds user interfaces and integrations, and tests them. The analyst only needs to describe the task in natural language. “In our approach, the no-code architecture ensures the system remains coherent and manageable, while AI agents reduce the effort required to build and enhance services,” said Nikolay Adeev, founder of Artsofte.

Banks can now use AI to build both internal information systems and remote banking platforms two to three times faster than with conventional no-code platforms. The developers have also addressed security by restricting the AI agents’ permissions: they have no direct access to banking databases or infrastructure, and all operations are performed through a secure gateway. This is particularly important in financial services, where requirements for network segmentation and personal data protection are significantly stricter than in the corporate software market as a whole.

The platform is especially relevant to smaller lending institutions that do not have in-house software development teams. It reduces their reliance on outside contractors, saves resources, and significantly speeds up the development of digital services. Integration between Abanking and the Agredator platform offers another important advantage. Under an agreement the companies signed in the spring of this year, banks gain a single connection point to key government data sources – the System of Interagency Electronic Interaction (SMEV), the Gosuslugi (Public Services Portal), the Pension Fund of Russia, the Federal Tax Service, and the Federal Bailiff Service – without having to build separate integrations with each agency.

Not Just in Russian

According to the Bank of Russia, one in five financial market organizations used artificial intelligence last year, while another third planned to adopt it within three years. Abanking AI Studio could help close that adoption gap: using the AI agents built into the platform requires neither capital investment nor an in-house team of programmers.

The platform supports multiple language models. Users can interact with its AI agents through Russian models such as YandexGPT and GigaChat, strengthening the country’s technological sovereignty. At the same time, international customers could potentially connect their own AI systems, expanding Abanking AI Studio’s export prospects. The company already works with international partners, and its no-code platform was deployed in Turkmenistan, Kyrgyzstan, and several African countries as early as 2024.

Levels of Access

Abanking launched its no-code remote banking platform in 2024 and moved into international markets almost immediately. At that stage, however, AI served only as an auxiliary tool. In 2025, T-Technologies’ team developed a fully fledged AI agent for programmers within the company’s ecosystem. It did more than generate snippets of code: it could also execute a sequence of specified actions. In 2026, FIS introduced AI agents for creating and modifying applications on FIS Platform, which is deployed at 10 of Russia’s largest banks.

AI agents can now work with data models, application functions, and software interfaces, reducing the effort required for individual tasks and accelerating service development.

A Shift in Roles

Routine tasks once handled by programmers can now be delegated to AI agents. But people remain responsible for designing complex system architectures, overseeing security, validating results, and managing quality.

The IT market is moving beyond AI as a coding assistant toward agentic systems capable of building software components on their own. Abanking AI Studio is another example of this trend. It helps level the technological playing field by giving regional players development tools that can match the speed previously available only to the largest banks.

The fact that generative AI can write code is no longer news. The real challenge is turning it from a convenient prototyping tool into a full-fledged development model within a large company, and integrating this new way of working into the production cycle, with all its requirements for architecture, security, and quality. We are now scaling this approach systematically because simply giving people access to AI is not enough. They also need to learn how to write prompts and switch between roles when interacting with the machine. We are already seeing results: in a corporate environment, AI can significantly accelerate the engineering cycle, reduce workloads, and free up time for architectural and product decisions

Svyatoslav Solovyev

Director of AI Department, Alfa-Bank IT