Why AI in Smart Buildings Gets Stuck, and What Operators Do Differently

Smart buildings have never generated more data. Energy and sub-metering, BMS and plant performance, occupancy and space utilisation, comfort and environmental conditions, maintenance history and helpdesk activity: most estates now collect far more than any facilities or workplace team could reasonably act on in a working week. AI is meant to close that gap, reading the building, surfacing what matters and telling operators what to do next.

And the pilots often go well. That is part of the problem. A proof of concept lands on one site, with one use case and one enthusiastic team, everyone can see the potential, and then it stops there. A year on, the estate has more dashboards, more vendors and more pilots, but the same decisions are still being made the same way they always were.

This session looks at why that happens, and at what the organisations who get past it do differently.

The harder problem is rarely the model or the sensor. It is the gap between buying smart building capability and giving the people who actually run the building enough confidence to use it, under time pressure, with an occupier on the phone. Facilities, workplace and estates teams are handed new tools and new dashboards without the data foundations, the operating norms or the decision frameworks that would let them trust what those tools tell them. So they verify it manually, or quietly fall back on the way that has always worked. Neither response is irrational. Both of them end adoption.