Almost every office wastes electricity in the same way. Not dramatically, and not because anyone is careless, but because a building that is occupied for roughly fifty hours a week is powered for a hundred and sixty-eight. Screens sit at standby, monitors glow, the coffee machine holds a boiler at temperature all weekend, and one meeting room air conditioner runs Friday night to Monday morning because somebody nudged it and left.
Nobody notices, because the bill arrives monthly as a single number with no story attached. The fix is not a heroic retrofit. It is a shutdown routine that runs at the same time every evening whether or not the last person remembers, and a start-up routine that has the place comfortable before the first person arrives.
Why manual shutdown never works
The reason is structural rather than human. Turning an office off is a task with no owner: the last person to leave changes daily, they are usually leaving in a hurry, and the reward for doing it properly is invisible. Any process that depends on the least-motivated person at the least convenient moment will fail most nights.
Automation changes the question from “did someone remember?” to “is the building empty?” That second question can be answered reliably by sensors, and acted on the same way every time.
Start with measurement, not switches
Before automating anything, find out what the building actually draws when nobody is in it. Fit a whole-building energy meter and look at the reading at three in the morning on a Sunday. That figure is your baseline load, and in most small offices it is between two and five times higher than people guess.
Then break it down. A handful of metering smart plugs rotated around the building for a fortnight will identify the offenders precisely, and it is nearly always the same short list: the kitchen, the meeting-room displays, the network cabinet, and one appliance nobody remembers buying.
Building the evening shutdown
A good shutdown is layered rather than a single event at a single time, because a hard 18:00 cut-off punishes the person working late and quickly gets disabled by whoever it inconveniences.
Layer one: the soft nudge
At around 18:00, dim the general lighting by twenty per cent and switch off decorative and display lighting. Nothing turns off that anybody needs. The change is deliberately noticeable, and it functions as a signal that the building thinks the day is over.
Layer two: the empty check
Fifteen to thirty minutes later, evaluate whether the building is genuinely empty. Do not rely on a single motion sensor, because someone sitting still at a desk registers as absent within minutes. The combination that works is mmWave presence sensors in the main working areas, which detect a motionless person, plus door contacts to confirm the last exit, plus a network check for how many devices are still associated with the Wi-Fi.
If all three agree the place is empty, run the shutdown. If any disagrees, wait and check again in fifteen minutes.
Layer three: the shutdown itself
In order: switch off meeting-room displays and projectors, kill the kitchen appliance circuit apart from the fridge, set heating and cooling to their setback temperatures, turn off all lighting except the statutory emergency and escape lighting, and finally arm the alarm. Announce the arming with an audible warning first, so anyone who was missed has a chance to intervene.
Layer four: the fallback
At a fixed late hour — 22:00 is typical — run the shutdown regardless of what the sensors say. If someone is genuinely still working, they will notice the lights change and can override it. A fallback matters because sensors fail, and a shutdown routine that quietly does nothing for three weeks because one sensor died is worse than no routine at all.
Building the morning start-up
Start-up is easier, because it does not have to decide anything. The one design decision worth making is to work backwards from the arrival time rather than forwards from a clock.
- Ninety minutes before the first arrival, bring heating or cooling from setback to comfort. Ninety minutes is a reasonable starting point for a well-insulated office; a draughty older building may need two hours or more. Measure how long yours actually takes and adjust once.
- Thirty minutes before, start ventilation so the building has exchanged its air before anyone breathes it.
- Fifteen minutes before, bring entrance and circulation lighting up, and power the kitchen circuit so the coffee machine is ready rather than warming up while three people wait.
- On first entry, disarm on a valid credential and let occupancy take over lighting for the rest of the day.
Two refinements make a real difference. Check a holiday calendar, so the building does not spend a bank holiday heating itself for nobody. And key the schedule to actual arrival patterns rather than official opening hours, which are frequently an hour apart.
What this is worth
We are not going to invent a percentage, because the honest answer depends entirely on your baseline, your tariff and your climate. What we can say is where the savings come from, in rough order of size.
The largest single win is almost always comfort systems running outside occupied hours — one air conditioner or heater left on over a weekend can exceed the entire lighting load for a week. Second is lighting in unoccupied areas, particularly circulation space and meeting rooms. Third is standby draw, which is individually trivial and collectively significant across forty devices. Fourth, and most easily overlooked, is reduced peak demand, which matters if your tariff includes a capacity charge based on your highest recorded draw.
Because you measured first, you will be able to state your own figure after a month rather than quoting someone else's.
Things that go wrong, and how to avoid them
The single motion sensor
A conventional passive infrared sensor detects movement, not people. Someone reading at a desk disappears from it within a few minutes. Either use presence sensors that detect a motionless body, or require several signals to agree before deciding the building is empty.
Switching something that needed to stay on
The classic failure is a smart plug on a fridge, a freezer or a network device. Do it once and you will lose a week's food or an evening's remote access. Decide up front which circuits are never automated, and write that list down where the next person will find it.
No manual override
Every automated space needs an obvious way for a human to say “not now”, and that override must expire on its own. An override with no expiry becomes permanent within a fortnight, and then the automation exists only on paper.
Nobody knows it exists
If the lights dim at six and nobody has been told why, the first assumption is a fault. Tell people what the building does and when. A routine that is understood is tolerated; one that is mysterious gets ripped out.
A sensible first month
Do it in this order and you will not have to undo anything.
- Week one: fit a building energy meter and read the overnight baseline. Change nothing.
- Week two: rotate two metering plugs to identify the biggest standby loads, and list the circuits that must never be automated.
- Week three: implement the evening shutdown, with a generous empty-check and the 22:00 fallback. Tell everyone.
- Week four: add the morning start-up, then compare the overnight baseline with week one.
That sequence costs very little, produces a number you can defend, and leaves you with a building that switches itself off every night without depending on anyone remembering.