Most automation systems today follow a reactive logic. Something has already happened, and the system tries to catch up.

It gets cold, so we add heat. It gets hot, so we switch on cooling. CO₂ levels rise, so we increase ventilation. At first glance, this seems like normal control logic. But in a real building, it often means something else: comfort conditions have already changed, people have already felt it, and the system has only just started to respond.

The problem with reactive automation

Building services systems rarely act instantly. This is especially true in large buildings with complex façades, water circuits, structures with high thermal inertia and rooms with different operating schedules.

If the controls wait until temperature or air quality has already moved outside the comfort range, they are always responding too late. And the larger the building, the more costly that delay becomes.

A building almost never responds instantly. There is the thermal inertia of its structure, air and water systems, the lag in people’s responses, and the influence of the sun.

From reaction to prediction

I believe that over the coming years, automation will gradually shift from reacting to predicting. Systems will look not only at what is happening now, but also at what is highly likely to happen in one, three or six hours.

This has become possible not because we have invented new physics, nor because some magical sensors have appeared. The reason is simpler: more data is available, computing has become cheaper, and building services systems have learnt to keep a record of their operation.

What the system needs to see

For a building to act ahead of events, knowing the current room temperature is not enough. It needs context:

  • weather forecasts
  • solar loads on façades
  • tenants’ operating schedules
  • actual room occupancy
  • the thermal inertia of the structure
  • a history of deviations, complaints and operating modes
  • electricity tariffs and power limits

The value is not simply in collecting more data. It is in linking that data to the right engineering decision.

My own home as a small-scale model

Some time ago, I decided to test this idea in my own home. I have a heat pump installed, and a simple question interested me: do we really have to wait until the house starts to become uncomfortable?

We began looking beyond the indoor temperature. We analysed the house’s thermal inertia, solar radiation, the time its structure took to heat up and cool down, and the system’s behaviour in different weather conditions.

It quickly became clear that the house did not behave randomly. It had its own memory, inertia and logic. Once you recognise this, decisions can be made before the temperature changes, rather than after.

Essentially, the question changed. Not ‘What is happening now?’ but ‘What will be happening to the building in a few hours?’

Commercial property will start predicting too

The potential in commercial property is even greater. Imagine an office building or shopping centre. The system knows tomorrow’s weather forecast, understands which façades will receive the highest solar loads, sees the tenants’ operating schedules and remembers how the building behaved in similar conditions before.

With this logic, the controls do not wait for the south-facing zone to overheat or for ventilation to start chasing rising CO₂ levels. They prepare the system in advance: changing operating modes, redistributing loads and accounting for peaks, tariffs and future occupancy.

In my view, this is where the greatest potential for the development of automation lies. Not simply in new sensors, controllers or attractive monitoring dashboards, but in a building’s ability to understand its own behaviour.

AI cannot compensate for poor engineering

At the same time, it is important not to turn this subject into yet another presentation about artificial intelligence. Predictive control only makes sense when it rests on sound engineering foundations: the right systems, well-understood hydraulics or air distribution, suitable sensors, proper configuration, commissioning and operating logic.

If the system itself is poorly designed, no algorithm will make it intelligent. It will simply reveal its weaknesses faster.

AI in a building only makes sense when it is based on sound engineering. Otherwise, it is simply an expensive way to make mistakes faster.

Decisions ahead of time

We have spent a long time teaching building services systems to react. Now we will gradually teach them to predict. I believe this transition will be one of the most important changes in building automation over the next decade.

Because the best system is not the one that quickly corrects discomfort. It is the one that prevents it from arising.

The most effective solution is the one that prevents the problem from arising at all.