My office routine is pretty predictable: I walk in, I usually know what I’m about to do, and I want the room to just… get out of my way. In this video I show how I built an adaptive office in Home Assistant that feels conversational and “AI-like” without needing an LLM. When I enter the office for the first time, Assist greets me, gives me a quick brief (including upcoming calendar events), asks what I’m doing, and then I answer with something like “work,” “video editing,” “filming,” “play a video game,” or even “just passing through.” From there, a script takes over and configures the room—lights, shades, temperature, and whatever else I’ve tied to that mode. The key is sentence triggers + carefully designed aliases. It takes a bit of upfront setup, but the payoff is a workflow that’s fast, local, predictable, and debuggable—exactly how I like my privacy-first automations. And if you do want extra flexibility, I also show an optional fallback where, if my canned responses don’t match, I hand the full response off to a local LLM to interpret intent. I walk through the script trace so you can see the exact decision paths, then I cover the hardware that makes it all work, including my SmartWave Matter shades (paired via an Apple TV border router) and the rest of my office stack.