Buildings consume a huge chunk of global energy and contribute significantly to greenhouse gas emissions. Optimizing how buildings operate is crucial for a sustainable future, and that’s where smart Building Management Systems (BMS) powered by AI come into play. Traditionally, creating these systems was a complex, building-by-building process. Imagine the time and effort involved in customizing models for each unique structure, each with its quirks and specific needs. This has been a major roadblock to wider adoption of smart BMS, creating a so-called *scalability problem*. Large Language Models (LLMs), the brains behind AI tools like ChatGPT, are stepping in to change the game. Researchers are exploring how LLMs can *automate the automation* of building energy management. This includes tasks like handling massive datasets, building predictive models, and tailoring these models to each building’s specific needs. The key innovation lies in *prompt engineering*. This involves crafting precise instructions to tell the LLM what to do—kind of like giving a highly skilled but literal-minded apprentice extremely detailed directions. The study tested three different *prompting* strategies: giving the LLM all the instructions at once (*one-shot prompting*), breaking them down into smaller, sequential steps, and a balanced, two-part approach (*bi-sequential prompting*). The results? The two-part, *bi-sequential* approach worked best. It allowed the LLM to generate accurate, functional code for a data-driven BMS while avoiding the pitfalls of too much or too little information. It also significantly outperformed a human beginner tackling the same task, completing in seconds what took a novice days. This approach achieved nearly perfect accuracy in both generating and executing code for building energy models, demonstrating its potential for other AI applications. While promising, the research also highlighted limitations. LLMs can still get lost if the instructions are too dense or too fragmented. There’s a sweet spot for communication—just like with human apprentices! Future research will focus on refining these prompts, making them smarter and more adaptable to different building types, and exploring other AI methods that can work in parallel to make the process even faster and more efficient. The vision? An AI assistant that can design the optimal energy strategy for any building, anywhere in the world, making our cities greener and our buildings smarter.