A chat interface using LLM is the new standard. It can be used not just for getting trivial or publicly available information, but also work on our company data to retrieve value out of it using language instead of typical and much more technical SQL language. In this talk we will delve into how AI agents can be leveraged to convert natural language queries into SQL and other programming languages code to make data analysis more accessible and less dependent on technical expertise. The presentation will also address the challenges of migrating from legacy systems, such as mainframes running Cobol, to modern platforms using LLM agents. We will demonstrate how AI can help reduce the risks associated with these migrations and ensure smoother transitions. Overall, during the talk we show where we are, the current challenges, and future ideas.