What makes a building smart? BMS, sensors and digital twins
How a smart building actually works — the control system that runs it, the sensors that feed it, the protocols that connect it, and why it pays off.
A smart building is not a building full of gadgets. It is one that measures its own conditions and adjusts its services — heating, air, light, access — without someone walking round to flick switches. The intelligence sits in the plumbing of the place: sensors reading the environment, controllers deciding what to do, and equipment that carries out the decision. When it works, you barely notice it. The room is the right temperature, the lights come up as you arrive, and the energy bill is lower than the square-footage would suggest.
This guide walks through the parts that make that happen: the control system that acts as the brain, the sensors that give it a picture of the building, the subsystems it drives, the protocols that let all of it talk, and the analytics that turn raw readings into decisions. It sits within our wider coverage of smart buildings, and it assumes no prior knowledge — every acronym is spelled out the first time it appears.
The building management system: the brain
At the centre of almost every commercial smart building is a building management system (BMS) — sometimes called a building automation system, or BAS. The two terms are used loosely and often interchangeably; BMS tends to imply a broader supervisory scope, BAS a narrower focus on automating plant. Either way, it is the layer that watches what the building is doing and tells the equipment how to respond.
A BMS is not a single box. It is a hierarchy. At the bottom sit field devices — sensors and actuators wired to the physical world. Above them sit programmable controllers, each responsible for a zone or a piece of plant, running the control logic that says, for example, “if this room is above 23°C and occupied, open the cooling valve”. At the top sits a supervisory layer with a screen a facilities team can use: floor plans, alarms, trends, schedules. The clever part is mostly in the middle. The controllers keep working even if the top layer goes offline, which is what you want from something that has to run a hospital or a data hall around the clock.
What the building senses
A control system is only as good as the readings it acts on. A modern building watches several things at once:
- Occupancy and presence — is a space in use, and roughly how busy is it? This drives lighting, ventilation and security, and it is worth understanding in its own right; our guide on occupancy and presence sensing covers how passive infrared, radar and other methods differ, and where each falls down.
- Temperature — measured per zone, and often at the supply and return of air-handling equipment, so the system can tell whether it is actually delivering what it promised.
- Humidity — too high and a space feels clammy and risks condensation; too low and it feels dry. Ventilation and cooling both affect it.
- Carbon dioxide (CO₂) — a useful proxy for how stuffy a room is and how many people are in it. Rising CO₂ tells the system to bring in more fresh air, which is the basis of demand-controlled ventilation.
- Energy and sub-metering — meters on electricity, gas, water and heat, broken down by floor, tenant or system. Without sub-metering you know the building’s total consumption but not where it goes, which makes waste almost impossible to find.
None of these sensors is expensive on its own. The cost and the value both come from wiring them together into a picture the control system can reason about.
What it controls, and what it senses
The BMS acts through the building’s subsystems. Each one has its own sensors and its own outputs. The table below maps the common ones.
| Subsystem | What it senses | What it controls |
|---|---|---|
| HVAC (heating, ventilation and air conditioning) | Zone temperature, humidity, CO₂, supply and return air, plant status | Boilers, chillers, pumps, fans, dampers and valves |
| Lighting | Occupancy, daylight level (lux), time of day | Switching, dimming and pre-set scenes per zone |
| Access and security | Door state, credential reads, motion, video | Locks, barriers, intruder-alarm arming, camera triggers |
| Energy and metering | Electricity, gas, water and heat use by circuit or area | Load shedding, demand response, tariff-aware scheduling |
| Fire and life safety | Smoke, heat, manual call points | Alarms, smoke dampers, stairwell pressurisation |
| Vertical transport | Lift car position, call demand, occupancy | Dispatch and grouping of lifts to cut waiting times |
Two honest caveats. Fire and life safety is usually a separate, independently certified system that the BMS may read from but is not allowed to override — you do not want a comfort algorithm second-guessing a smoke detector. And HVAC is where most of the energy and most of the complexity live, which is why so much smart-building effort is really HVAC effort wearing a nicer interface.
Speaking the same language: protocols
Here is the recurring headache. The equipment in a building comes from many manufacturers, bought over many years, and it does not naturally speak a common language. Several protocols try to bridge that gap.
- BACnet (Building Automation and Control network) is the most common open protocol in commercial buildings, especially for HVAC. It defines standard ways to describe a sensor, a set-point or a schedule so that gear from different vendors can, in principle, interoperate.
- KNX is an open standard more common in Europe and often used for lighting, blinds and room control, running over a shared bus.
- Modbus is older and simpler — a plain, widely supported way to read registers from meters, drives and industrial devices. It carries data but says little about what the data means.
“In principle” is doing a lot of work in that first bullet. Two devices can both claim to support a protocol and still refuse to cooperate, because one vendor implemented an optional feature and the other did not, or because the data is technically present but labelled in a way nobody standardised. Integrators spend real time mapping one system’s “AHU-3 supply temp” onto another’s cryptic register address. Interoperability on paper and interoperability on site are different things, and closing the gap is a large part of what a controls contractor is actually paid for.
From readings to decisions: analytics and digital twins
Once the data is flowing, the next step is making sense of it. Basic building management shows you live values and raises an alarm when something crosses a threshold. Analytics goes further, looking for patterns a human scanning dashboards would miss: a chiller and a boiler fighting each other, a zone that never reaches set-point, a fan running full tilt at 3 a.m. in an empty building. This kind of automated fault detection and diagnostics turns a flood of readings into a short list of things worth fixing.
The richest version of this idea is the digital twin — a continuously updated virtual model of the building, fed by its live sensor data. A twin lets you ask “what if” questions without touching the real plant: simulate a heatwave, test a new control strategy, or check how a maintenance shutdown will ripple through the rest of the systems. The term is used loosely, and a static 3D model with a few live labels is sometimes sold as one. A genuine twin is defined by the feedback loop — the model and the building stay in step, each informing the other.
In short: a building becomes "smart" when a management system links its sensors (occupancy, temperature, humidity, CO₂, energy) to its subsystems (HVAC, lighting, access) through shared protocols such as BACnet and KNX, then uses analytics — and sometimes a digital twin — to run everything with less energy, more comfort and fewer surprises.
What it is all for
The point of the exercise is a handful of measurable outcomes.
Energy. HVAC is typically the largest single energy load in a commercial building, and lighting is not far behind. A system that ventilates and cools spaces only when they are occupied, and that catches equipment running against itself, cuts consumption that would otherwise be invisible on a single monthly bill.
Comfort. Steadier temperatures, better air quality and light that suits the task keep occupants comfortable. That is not a soft benefit: in an office, staff cost far more than energy, so a building that keeps people alert and un-distracted pays back quickly.
Maintenance. The oldest model is run-to-failure — fix it when it breaks. Time-based servicing improves on that but wastes effort on equipment that was fine. Condition-based and predictive maintenance use the sensor data to service things when the data says they need it, catching a failing bearing or a clogging filter before it becomes an emergency call-out on a Friday night.
Retrofit versus new build
How you get there depends heavily on where you start.
A new build is the easy case, relatively speaking. Sensors, cabling, controllers and a coherent BMS can be designed in from the outset, chosen to work together, and commissioned before anyone moves in. The building’s structure and its intelligence are planned as one thing.
A retrofit is the common case, because most of the buildings that will exist in twenty years already exist today. Here you are working around old plant, incomplete drawings and systems that were never meant to talk to each other. Wireless sensors and gateways help, because they avoid tearing open walls to run new cable, and a modern supervisory layer can often sit on top of ageing controllers and translate between protocols. The trade-off is that you inherit the building’s quirks. Retrofits live or die on a good survey up front: knowing what is actually installed, what still works, and what is quietly lying about its own status.
Where to start
If you are weighing up whether a building is smart, or could be, the useful questions are practical ones. Can you see energy use broken down by system, not just as one number? Does the ventilation follow the people, or the clock? When something fails, does the building tell you before the occupants do? A building that answers those well is doing the job, whether or not it carries the label. Metering and clear data tend to be the highest-value first steps, because you cannot manage — or automate — what you cannot yet see. Everything more ambitious, up to a full digital twin, is easier to justify once the measurements are in place and someone is actually reading them.