The article explores the potential of using multi-agent AI systems in the banking sector, particularly for automating business processes and improving operational efficiency. It emphasizes the growing importance of integrating artificial intelligence (AI), machine learning (ML), and agentic architectures, especially in high-stakes, regulated environments. The authors identify a research gap in evaluating combined applications of multi-agent systems and large language models (LLMs) in banking. The study uses qualitative research methods, including literature analysis, IT solution reviews, case study examination, and interviews. The research focuses on implementations in Credit Agricole Polska, examining two existing AI tools (“WALL-E” and “ASIA”) and modeling their integration with a hypothetical agentic system. Business processes before and after AI integration were mapped to identify efficiency improvements. The research showed that applying multi-agent systems could significantly streamline operations like document processing, client communication, and customer support. The WALL-E system reduced delays by automating document classification and routing, while ASIA improved customer service through AI-driven chatbot interactions. The authors mapped processes which are operated by an IT system integrating AI and those which possibly could be operated by multi-agent systems. The paper discusses challenges such as system integration, regulatory compliance, and ethical concerns in AI deployment. It recommends phased, modular implementation and emphasizes the need for cross-functional teams to ensure transparency, accountability, and scalability. Limitations include the focus on a single institution and the conceptual nature of the proposed model.