Automating a building: what pays for itself
Some automation returns its cost in a year. Some is an expensive light switch. An honest split — and the rules that keep it working when the grid does not.
Building automation is sold as a lifestyle and bought as an investment, and the mismatch is why so many installations end up half-used. The lights change colour for a month, then someone flips the wall switch and never opens the app again.
The projects that stay in use are the ones that removed a cost or a risk, not the ones that added a feature. Here is the honest split, from systems we have installed and still support.
What usually pays for itself
- Scheduling heavy loads. Air conditioning and water heating are the bulk of a building's consumption, and both are routinely run for hours in empty rooms. Automation tied to real occupancy and working hours — not to a hopeful timer — is the single most reliable saving available.
- Measuring energy at all. Most buildings cannot say which floor, which tenant or which machine consumes what. Meters and a dashboard rarely reduce anything by themselves, but they end the arguments and they make every later decision evidence-based. This is usually the cheapest step and the highest-leverage one.
- Access control instead of keys. Keys get copied, lost and never returned; someone has to be at a desk to hand them over. Replacing that with cards and codes removes a recurring labour cost and a permanent security hole — and it produces a log, which keys never did.
- Monitoring with alerts. A cold room drifting warm, a pump that stopped, a water tank running dry, a generator that failed to start. The value here is not comfort, it is the loss that did not happen. One prevented spoilage event often exceeds the cost of the sensors.
- Switching sources cleanly. In a building that moves between grid, generator and battery several times a day, automating the transfer — and recording it — protects equipment and tells you what the real availability actually is.
What usually does not
- Motorised curtains everywhere. Wonderful in the showroom. In practice, the ones people actually use are in one or two rooms; the rest are an expensive way to do what a hand does for free.
- Voice control as the primary interface. It struggles with mixed Arabic and English commands, it fails in a noisy room, and it is useless to a guest who does not know the phrases. Fine as a bonus, poor as the plan.
- An app per device. Six brands, six accounts, six apps that each work alone. This is the most common way a "smart building" becomes a collection of gadgets nobody controls as a system.
- Anything that only works with internet. Which brings us to the rule that matters most here.
The rules that decide whether it survives
The wall switch must always work
If the only way to turn on a light is a phone, you have installed a problem. Every automated circuit needs a physical control that works when the network is down, the phone is flat, or the person standing there is a guest. Automation should be the fast path, never the only path.
Decisions happen locally
A controller that has to ask a server in another country whether to open a door is a controller that fails when the link does. Local rules, local decisions, and the cloud used for reporting and remote access rather than for permission. This is exactly how we built our own access device: the door decides at the door, and syncs its log upward when connectivity returns. We have written about that principle more broadly in power cuts are a design input.
Design for the transfer, not for the outage
The damaging moment is rarely the outage itself — it is the switch between sources. Devices reboot, and anything mid-operation is interrupted. Controllers, the network switch and the access points belong on protected power, or you get a system that restarts several times a day and blames the software for it.
Wiring is permanent, logic is not
You can change a schedule in ten seconds and a cable in ten days. During construction or renovation, run more cable than you need to more positions than you plan to use — the marginal cost then is trivial, and retrofitting later is the entire budget. If the building is already finished, plan around wireless honestly rather than pretending the walls are not there; the same physics applies as with Wi-Fi coverage.
Industry: measure before you automate
On the industrial side the sequence that works is almost always the same: instrument first, watch for a while, then automate what the data proves is worth automating.
Sensors, meters and controllers feeding a live dashboard — temperatures, power, machine states — with alerts before small problems become stoppages. Two or three weeks of that will usually reveal that the expensive problem is not the one everyone assumed. Automating a process you have not measured is how you make a bad process faster.
Questions to ask before signing
- What still works if the internet is down for a day? Show me.
- What still works if the vendor disappears in three years?
- Can I operate every automated device manually, and where is that control?
- Is this one system, or several apps sharing a wall?
- What exactly is measured, and can I export it?
- Who services it, how fast, and what are the spares?
- What does the design cost, and do I own it if I do not proceed?
How we approach it
Every automation project starts with a site visit and a written design — devices, wiring and exactly how each room or production line behaves — with a fixed price attached before any equipment is ordered. Then installation, then a walkthrough where every scene, sensor and fail-safe is demonstrated to you rather than described.
We build the hardware as well as the software, which matters more than it sounds: when off-the-shelf does not fit, the answer is a device designed for the job — board, casing, firmware and the local decision logic that keeps working when the network and the power do not. That is the difference between a building that responds and a building full of gadgets.