How We Rebuilt Our Chatbot with Agentic LLMs
When we launched our GenAI chatbot for customers in April 2024, it was built on state-of-the-art technology. But progress in GenAI has been rapid, and by late 2025, language models had improved dramatically to a state where they can reliably work agentically in a production setting. Agentic AI allows for faster response times, better modularity and maintainability, a smoother user experience and the ability to handle more complex situations and requests. For this reason, we made a substantial investment: completely rebuild our (1.5 year old) legacy code around the new paradigm of agentic AI. In this talk, we will explore what agentic AI actually means (as opposed to marketing and buzzwords), and how we redesigned our entire architecture around it. And of course, we will demo our new agentic chatbot to show how it improves the holy grail of frontend products: the customer experience.
