Learning From Building Behaviour
How Automatry is thinking about machine learning, building behaviour, and engineer-led optimisation without removing human judgement.
A careful look at using models to support better evidence, earlier insight, and future optimisation work.
A good engineer can often tell when something is wrong by looking at a trend, a valve position, a room temperature, or a plant response. The aim of our machine learning work is not to replace that judgement. It is to support it with better evidence and faster insight.
We are exploring how Edge can learn normal building behaviour, identify drift, and highlight systems that deserve attention. The useful output is not a mysterious score. It is a practical signal that helps an engineer understand where to look next.
Learning before controlling
Optimised control is a long-term direction, but the first step is understanding. Before any system can recommend a better operating pattern, it needs reliable data, asset context, safety boundaries and a clear view of what the building is already doing.
That is why our work starts with visibility, anomaly detection, commissioning evidence and engineer review. Better models are only useful if they sit inside a trusted process.
Engineer-led optimisation
The goal is to help teams tune buildings more intelligently, not to remove accountability. Any meaningful optimisation layer needs safeguards, audit trails and human oversight, especially when it touches comfort, energy and control strategy.
Edge gives us the foundation for that work: live data, structured assets, operational context and a route from insight to action.