Introducing AI into engineering workflows is more complex than it appears. Over the past several months, I’ve led an initiative to integrate agentic AI into our R&D processes in Next Insurance. This journey involved thoughtful planning, experimentation, and team engagement to ensure a meaningful impact while minimizing security risks. I'll share our approach—from strategy and rollout tactics to overcoming human and technical barriers. Through this talk you’ll gain insights into what worked, what didn’t, and what we learned along the way.