In an episode of the High Agency podcast, Peter Gostev, head of AI at Moonpig, discusses his experiences with implementing large language models (LLMs) at Moonpig and previously at NatWest. He reflects on the challenges and strategies involved in deploying AI within large organizations, emphasizing the need for experimentation and a diverse portfolio of projects. Gostev shares insights on the excitement and unpredictability of new AI developments, such as real-time voice APIs and fine-tuning for vision models, and highlights the importance of curiosity in leveraging AI tools effectively. He also notes that small, agile teams can achieve meaningful improvements in productivity and cost savings, even if significant revenue impacts are still forthcoming. The conversation underscores the rapid evolution of AI and the potential for underappreciated advancements like OR1 models, encouraging AI leaders to embrace a hands-on approach and continuous learning.