AI agents are becoming increasingly prevalent as digital assistants capable of monitoring tasks, making decisions, and executing actions autonomously. The architecture of AI agents typically includes a Large Language Model (LLM) for understanding and processing language, tools for interacting with external systems, and a reasoning loop for problem-solving. The growing availability of low-code frameworks simplifies the creation of these agents, enabling users to build agents with basic configurations. The text describes a project to develop a fintech AI agent from scratch using Python, focusing on loan and insurance decision-making. It involves creating three specialized agents: a Loan Agent employing machine learning to assess loan eligibility, an Insurance Agent using vector databases for claim validation, and a Kernel Agent to direct queries appropriately. The flexibility of raw Python allows customization beyond black-box frameworks, enabling businesses to integrate their own data and decision-making criteria. The implementation illustrates potential efficiency gains for fintech companies in automating loan and insurance evaluations, reducing manual labor, and increasing decision reliability.