The emergence of accessible generative artificial intelligence models, notably ChatGPT in 2022, has sparked a new wave of digital transformation in the banking sector. Banks around the world have begun experimenting with GAI, calling it an “explosion of innovation” that has the potential to expand the industry’s capabilities radically.

Generative AI is a type of technology based on large language models (LLMs) that can create content: human language, code, images, etc. In the financial sector, this technology opens new avenues for improving operational efficiency, customer service, and data-driven decision-making.

In this review, the MIM:AGENCY team revealed the main areas of application of generative AI in the banking sector, examples of real implementations in the US, EU, and Ukraine, key benefits, as well as risks, challenges, and the state of technology development in Ukrainian banks.

Main areas of application of generative AI in banks

  • Customer service and chatbots. Generative chatbots are capable of conducting dialogues in human language, understanding the context of questions, and providing detailed personalized answers. For example, ING Bank has developed a chatbot that, in its pilot phase in 2023, was able to automatically resolve 20% more requests, reducing the need to wait for an operator.
  • Personalized financial offers. With LLM, banks can analyze customer data more deeply and offer customized products. Morgan Stanley is introducing a GPT-4-based chatbot for its financial advisors that searches for information in the bank’s vast internal research library. In Ukraine, OTP Bank analyzes the emotional tone of a customer’s voice during a call and generates personalized offers.
  • Business process automation. ABN AMRO Bank is piloting the use of ChatGPT to automatically summarize conversations between call center employees and customers. Goldman Sachs is experimenting with AI-based code generators that help automate up to 40% of typical code snippets.
  • Fraud detection. GSI is capable of generating examples of fraudulent transactions to improve the training of monitoring systems. JPMorgan Chase uses a ChatGPT-like tool to analyze employee emails for signs of financial fraud. Mastercard has used generative AI to search for compromised cards, doubling the speed of detection and dramatically reducing false positives.
  • Credit scoring. Models can generate understandable explanations for credit denials to customers, create synthetic borrower data to test the reliability of scoring models, and analyze non-standard information during the underwriting process.
  • Financial analysis and forecasting. Bloomberg introduced its own LLM BloombergGPT, which outperformed general models in financial analysis and sentiment analysis of market communications. Banks use GSI to forecast financial indicators and generate various development scenarios.

Key benefit

  • Increased productivity. SouthState Bank reported that employees are replacing about 50% of search queries with an internal AI bot. Tasks that previously took 12-15 minutes are now completed in seconds. Deloitte estimates that the use of GAI could increase the productivity of banks’ front offices by 27-35% by 2026.
  • Improved customer service. Customers receive more detailed and relevant answers to their questions around the clock, which increases their satisfaction with the service.
  • Better quality decisions. AI models help make more informed decisions by providing additional insights and synthesizing vast amounts of knowledge.
  • Reduced operating costs. The widespread adoption of AI can reduce the banking sector’s costs for performing individual operations by 20-25% by eliminating duplication and improving accuracy.

Key risks and challenges

  • The problem of “hallucinations.” Generative models can sometimes give false or fabricated answers, presenting them with excessive confidence. The probability that AI will answer “I don’t know” is almost zero – instead, it tends to make something up and even generate fake links.
  • Data privacy and cybersecurity. Using third-party AI services is potentially dangerous because internal data may end up on external servers. Many banks, including JPMorgan, have temporarily banned employees from using ChatGPT at work.
  • Model bias. Generative models are trained on historical data, which may contain hidden biases. There is a risk that AI solutions in the lending sector may discriminate against certain groups.
  • Regulatory restrictions. Many regulations do not take into account the nuances of AI use. The European Union is preparing to adopt an AI Act that will require AI decisions to be explainable, non-discriminatory, and subject to external audit.
  • Technical challenges. Large models require powerful computing resources, and the lack of personnel with the necessary expertise is a significant problem.

State of implementation in Ukraine

The Ukrainian banking sector has historically been quite technologically advanced – Ukrainian banks were among the first in the world to introduce contactless payments, online banking, and digital documents. So it is not surprising that they began to master artificial intelligence quite early. Even before the era of generative AI, Ukraine’s largest financial institutions used machine learning for scoring, financial monitoring, and first-generation chatbots. According to the Association of Ukrainian Banks, AI technologies have already been implemented by banks such as PrivatBank, Oschadbank, OTP Bank, PUMB, Sense Bank, Universal Bank (monobank), and others.

First and foremost, this concerned the automation of business processes, risk assessment, and marketing personalization. In particular, ML-based scoring systems are now in use at almost every major bank – they predict the probability of borrower default based on numerous parameters and quickly segment customers for product offerings. Some banks (e.g., PUMB) have chatbots on Viber/Telegram for basic consultations, which also use elements of NLP.

As for generative AI, mass interest in it in Ukraine awakened after the success of ChatGPT. In 2023, several banks announced pilot projects in this area. As already mentioned, PrivatBank has launched an AI transformation program at the corporate level. Within its framework, an AI Center of Excellence and an AI Governance system have been created, and a number of specific cases have been launched: an AI chatbot for serving business clients, internal AI solutions for optimizing processes in departments, automation of routine operations, etc.

In fact, the country’s largest bank is betting on the comprehensive implementation of AI in all areas, from front office to back office. Oschadbank, in partnership with the Ministry of Digital Transformation, held a hackathon in 2023 to develop AI solutions for the bank, seeking talent and fresh ideas from startups. Raiffeisen Bank (Ukraine) has also announced its intention to integrate AI into the small business credit analysis process. Monobank (Universal Bank) is known for its digital innovations: its co-founder, Oleg Gorokhovsky, has repeatedly mentioned on social media experiments with GPT to improve the customer experience. For example, the idea of AI analysis of customer transactions with savings tips or an AI assistant in the support chat was discussed. These ideas are still in the experimental stage, but it is likely that they will be implemented in the coming years.

The potential of the Ukrainian AI market in banking

The potential of the Ukrainian AI market in banking is significant. First, Ukrainian banks can more quickly borrow best practices from global players, avoiding their mistakes. For example, seeing the results of Morgan Stanley or ING projects, our banks can adapt similar solutions for themselves. Second, Ukraine has a highly developed IT industry, many AI startups, and talented people. This means that banks can collaborate with local AI companies or hire strong specialists for in-house development. 

There are examples of fintechs already working on AI products for banks – for example, Scorto offers AI platforms for scoring, and BIS Soft offers chatbots for banks with AI elements. Third, high digital literacy among customers: users are accustomed to apps such as Diya and Monobank and will be open to interacting with smart bots or AI services. This creates demand from below. At the same time, as everywhere else, the effect of AI implementation will depend on trust – whether customers will trust decisions that have AI “under the hood.” Therefore, banks need to communicate these things transparently.

So, Ukrainian banks are at the beginning of the road to implementing generative AI. Some have already taken the first steps (chatbot pilots, internal projects), while others are still watching. But it is clear that over the next 2-3 years, we will see significantly more AI initiatives in Ukraine’s financial sector. According to a survey of experts, Ukrainian banks are even more enthusiastic about implementing customer-oriented AI solutions than their foreign counterparts, which could become a competitive advantage for the local market. The potential for using GSI – from increasing efficiency to creating new products – is enormous, so those players who are the first to master this technology will gain a significant market advantage.

Conclusions and recommendations

Generative artificial intelligence is gradually transforming from a novelty into an integral part of banks’ toolkit. Its ability to understand and formulate human language, learn from large data sets, and generate creative solutions opens up new horizons for financial institutions to increase efficiency.

The experience of banks in the US and Europe has shown that GAI can significantly improve both internal processes (automation, analytics, fraud prevention) and customer interaction (service personalization, speed of service). The first implementations have brought measurable benefits, from time and money savings to increased customer satisfaction and the emergence of innovative services. Despite the challenges of wartime, Ukrainian banks are not standing aside from this trend and have every chance to use GSI to leap forward in their digital transformation.

At the same time, when implementing generative AI, it is critically important to be aware of the risks. Improperly controlled AI can damage a bank’s reputation or even lead to financial losses and penalties. Therefore, it is recommended to adhere to the “human-centric” principle: AI should be a tool to assist, not a complete replacement for human decisions. At each stage, there should be an opportunity for verification and intervention by a specialist. It is also worth introducing Responsible AI policies and an internal control system for AI models in the bank.

Key recommendations for banks on implementing generative AI:

  1. Start with small pilots and internal cases. Select 1-2 priority areas where generative AI can have a quick impact (e.g., an internal chatbot for IT support or report generation) and implement a pilot project. This will allow you to gain experience with minimal risk and demonstrate quick wins. Successful solutions can then be scaled to other departments.
  2. Invest in data quality and infrastructure. Ensure that the data on which the model is trained and operates is free of noise and bias. Establish a process for updating datasets. Provide a secure environment for deploying models (servers, cloud solutions, with the necessary security certificates). This is the foundation without which AI will not deliver the desired results.
  3. Create a cross-functional AI team. Bring together specialists from IT, data analysis, business units, and risk management to work together on AI projects. Such a team must understand both the technical aspects and the business context of AI application. Consider appointing a person responsible for AI (Chief AI Officer or Head of AI), as leading banks do.
  4. Provide staff training. Introduce training programs for employees on how to use AI tools in their work, how to interpret their results, and what to pay attention to. A culture of trust in new technologies is formed when people understand how they work. In particular, teach front-office employees to work correctly with AI so that they do not blindly trust models but critically evaluate their assistance.
  5. Pay attention to cybersecurity and privacy. Along with the introduction of new AI services, update your security policies: who has access to the models, what data can be entered, and how the model’s actions are logged. Make sure that only authorized data sets are used and that there is no risk of confidential information being leaked. Conduct stress testing of AI systems for possible attacks and abuse.
  6. Communicate changes to customers and regulators. If you are launching a customer AI service (e.g., a chatbot), inform customers how to use it and what issues it can resolve. Emphasize that this is an innovation for their convenience. At the same time, maintain dialogue with the NBU: share your experience, learn about their vision, and possibly join AI working groups. Transparent communication will increase trust and alleviate potential concerns.
  7. Don’t wait, take gradual action. Generative AI is a rapidly developing technology. There is already a risk of falling behind if you delay its adoption. So it’s worth taking a proactive stance: experiment, learn from the mistakes and successes of others, and look for your own niches for AI applications. At the same time, act responsibly and prudently, being aware of the risks. As noted at the Money20/20 conference, this technology “unleashes innovation in areas that we can’t even think about”. Ukrainian banks need to become part of this global wave of innovation.

Sources:

  1. Reksoft. The use of generative AI in banking. Analytical report. September 2025.
  2. (Main document containing case studies from ING, ABN AMRO, Morgan Stanley, SouthState Bank, Mastercard, Wells Fargo, JPMorgan Chase, Goldman Sachs, Deloitte, etc.)
  3. Money20/20 Europe Conference Reports (2023–2024).
  4. Deloitte Insights. Generative AI in Banking: Productivity Scenarios and Financial Impact. Estimates of front-office and back-office productivity as a result of GSI implementation.
  5. Bloomberg. BloombergGPT: A Large Language Model for Finance. Research on LLM trained on financial data and its application in market analysis.
  6. European Central Bank (ECB). Artificial Intelligence in the Financial Sector: Opportunities and Risks. Analytical materials on the impact of AI on the regulatory regime and supervisory activities.