AI won't fix bad building data.

And that's something the smart buildings industry needs to talk about more openly.

Right now, there is enormous excitement around AI in building operations — predictive maintenance, autonomous optimisation, digital twins. But there's a practical challenge hiding underneath all of that.

Most Building Data isn't AI-ready.

Commercial buildings generate huge amounts of operational data every day from:

• BMS systems
• HVAC equipment
• Energy meters
• IoT sensors
• Lighting and security systems

But the reality inside many buildings looks like this:

• Inconsistent point naming (TEMP1, AI-23, SA_TEMP)
• Missing or unreliable sensor data
• Systems locked behind proprietary vendor platforms
• Data with little or no contextual metadata

In other words, a lot of the data is messy.

And AI models are only as good as the data they are trained on. If the inputs are inconsistent, incomplete, or poorly structured, the outputs will be unreliable.

This is why the first step toward AI-powered buildings isn't actually AI.

Step one on any AI journey is building a clean, structured data foundation.

Buildings and Building Data needs context.

The second reason this problem has been historically difficult to solve is that it sits at the intersection of software engineering and building operations.

Many pure technology teams have attempted to tackle building data problems over the years. They often bring strong technical skills in AI, analytics, and software architecture. But building systems are highly specialised environments with decades of legacy equipment, unique protocols, and operational nuances.

Understanding what building data actually means — and how systems behave in the real world — requires deep domain knowledge of HVAC, controls, and building operations.

Building the team that combines both software engineering and building engineering expertise takes four to five years.

That's one of the reasons so many attempts to solve building data problems with technology alone have struggled.

Even if AI eventually helps address some of these challenges, most organisations will still need teams that understand both the technology and the buildings themselves to successfully implement it.

Summary

Standardising building data, integrating systems, and organising metadata may not sound as exciting as AI, but it's the work that makes AI possible.

This is exactly why we've spent so much time at Switch Automation focusing on the data layer in buildings — integrating systems, standardising metadata, and organising operational data across portfolios.

It's not always the most visible work in smart buildings, but it's the foundation that makes advanced analytics and AI actually useful in the real world.

And one of the reasons the team at Switch Automation is so valued by our clients is the real-world experience and domain expertise we bring — helping de-risk projects and ensuring they stay on time and within budget.