Small automations a family actually loves

Smart-home demos are full of colour-changing lights and phone apps with fifty toggles. Nobody in a real family uses those after week one. What people love — what they mention to friends — are the small, considerate things the house does without being asked, and gracefully stops doing the moment a human wants something different. These are the ones running in our house today, and why each one is shaped the way it is.

What makes an automation lovable

Three rules, learned from years of daily family use. Invisible until useful: a good automation is noticed only when it saves you a trip across the room, never when it interrupts. The human always wins: if you do something by hand, the house treats that as the final word — it never fights you to restore its own idea of correct. Honest about what it did: when the house acts or answers, it verifies the result first, and admits plainly when something didn’t work. An assistant that occasionally says “that didn’t start, I’ll not pretend it did” earns more trust than one that is cheerfully wrong.

The ones that earn their keep

Movie night handles the curtains. Turn on the living-room TV and the curtains glide closed. Turn it off, and they return to exactly where they were — if they were sitting at a third open for the afternoon light, that is where they go back to, not just “open”. And if someone opens them mid-film, the house takes the hint and stands down. Why it’s shaped this way: restoring the room how you had it — not how the system defines “normal” — is the difference between a house that serves you and one you serve.

“Turn the aircon off in twenty minutes.” The assistant takes instructions for later: lights, aircon or curtains, any room, at a time or after a delay, in English, Thai or Mandarin. It replies with the verified time it actually scheduled (“I’ll close the curtains at 21:40”), you can ask what’s pending, and you can cancel by voice. Schedules clean themselves up after firing — no orphaned timers haunting the system. Why it matters: households run on “later”, not “now” — falling asleep to the fan, curtains at sunrise, the kettle-and-school-run choreography. An assistant that can’t take an instruction for later is only half an assistant.

Lights that know dark from daytime. Walk into the cooking room and the lights come on — but only when the room is actually dim, because the motion sensor carries its own light meter. In the shower, millimetre-wave presence sensing knows you are there even when you are perfectly still, so the lights never plunge you into darkness mid-shampoo. Everything switches itself off a few minutes after the room is truly empty. Why it’s shaped this way: motion is not presence, and “someone moved” is not “someone needs light”. Getting those two distinctions right is most of the difference between delight and annoyance.

The house knows the cleaner comes on Friday. When the last family member leaves, the house switches off the lights and air conditioning behind you. But on Friday mornings it deliberately stands down — because that is when our cleaner is here, and nobody wants the aircon dying around the person doing the ironing. It re-arms itself on Friday evening without anyone remembering to. Why it matters: real households have helpers, guests, routines and exceptions. An automation that ignores the household’s actual rhythm gets switched off within a month; one that respects it becomes part of the family.

A welcome tuned to each person. The house notices who has come home and does the right thing for that person — the bedroom aircon starts cooling for one of us, the living-room lights come up for others. Arrivals are personal, not generic. Why it’s shaped this way: a home is not one user account. Everyone in the family — three generations, three languages — should feel the house was set up for them personally.

A camera that guards the house, never the family. The kitchen camera records only when the house is empty. The moment anyone comes home it switches off — and it is blocked from switching on while the system starts up, so there is no gap where it could catch the family by accident. Recognition of faces happens on our own hardware; no footage ever leaves the building. Why it matters: this is the question every family quietly has about smart cameras. Answering it in the design — provably, not as a promise — is what makes cameras acceptable in a home at all.

The house looks after its own machines. When Bangkok’s hot season pushes the server drives too warm, the house turns on the living-room aircon, waits for the drives to cool, then puts the room back the way it found it — and sends a message if the heat won’t break. The full guide is on the hobby page. And when the storage box comes up a minute after everything else, the services that depend on it wait for it rather than starting on empty disks — the mount guard is the newest guide. Why it’s shaped this way: escalate gently, restore politely, and only involve a human when the automatic answer has genuinely run out — the same pattern as every automation above, applied to the machines themselves.

“Where’s Mum’s phone?” Ask, and the right person’s phone rings out loud — the assistant knows whose phone is whose, tolerates the microphone mishearing names, and works in all three languages. The house can also answer who is home, and greet family members by name at the door using face recognition that runs entirely on our own hardware. Why it matters: these are the questions families actually ask, dozens of times a week. Solving the unglamorous ones well is what makes the system feel like part of the household.

Why we build them this way

Every automation above follows the same handful of patterns, and we apply them deliberately. Snapshot, act, restore: before the house changes anything on your behalf, it remembers how you had it, and puts it back — verified, not assumed. Manual override wins: a hand on a switch or a curtain always outranks the automation, silently and immediately. Suppression windows: the schedule of a real household — the cleaner’s day, the weekend lie-in — is part of the logic, not an afterthought. Deterministic first: everyday actions run on fixed, predictable logic that behaves identically every time; the AI layer sits above it for requests that genuinely need understanding. And verify before claiming: the house checks that the light went off, the curtain moved, the schedule exists — before it says so.

This is also, candidly, the pitch to the professionals we work with: installing devices is the easy half of home automation. Understanding what a family will still be using — and still trusting — a year later is the hard half, and it comes from living with these systems, in a three-generation, three-language household, every day.

Where the voice fits in. Most of these automations can also be driven, scheduled or interrupted by speaking to the house in Thai, Chinese or English — the voice layer has its own tour, and the whole system is what our AI Hub productises.