What happens when AI answers your DMs at 02:00
A customer never decides to message you at an inconvenient time. She decides she wants something — and reaches for her phone. Sometimes that's 14:00. Sometimes it's 02:00.
Night messages don't wait for opening hours
Here is a small, ordinary loss that happens to studios and cafés every week. At 02:00 someone can't sleep, opens Instagram, and asks a beauty studio whether there's a facial free on Saturday morning. It's a warm question — she's ready to book. But the studio is closed, the owner is asleep, and the message sits there.
By the time the owner opens the inbox at 09:00, the customer has already found a studio that answered. Nothing dramatic happened. No one complained. The booking simply went somewhere else, quietly, while everyone who could have caught it was off the clock.
You can't ask an owner to watch a phone until 02:00. What you can do is make sure a clear, honest answer is waiting when the question arrives — and that the owner still gets the final say in the morning.
What the assistant actually does overnight
The SmashOne assistant isn't a personality and it isn't a guess engine. It answers from things you've already written down: your business profile, your FAQ, and your catalogue — prices, opening hours, and what's available. If a café lists a flat white at €3.40 and a closing time of 18:00, that's exactly what the assistant tells people. Nothing invented, nothing improvised.
- It answers from your profile, FAQ and catalogue — prices, hours and availability you've set yourself.
- Every reply carries a short, clear notice that it's an assistant, not a person — so no one is misled.
- Voice questions are understood too, and each counts as one reply, the same as a typed message.
- It never invents an answer. If it isn't sure, it says so plainly and takes a message for you.
That notice isn't a footnote we hide. Under European AI-transparency rules a person has the right to know when they're talking to a machine — so the label is on the reply itself, in plain language, every single time.
When it wakes you up — and when it doesn't
Some questions should never be answered by a machine, and the assistant knows the difference. Billing, legal and refund questions always go to a human — they're held for you, not guessed at. And when a message reads as upset, sensitive, or simply too complex, the assistant doesn't push. It says the owner will get back to you in the morning, and adds the conversation to your review queue.
So the night splits cleanly. Straightforward questions — price, hours, what's free on Saturday — get a real answer immediately. Everything that needs judgement waits, clearly flagged, for the first cup of coffee. Here's roughly how an overnight exchange looks. It's illustrative — a generic studio, not a real customer.
Passed to the owner. Refunds are handled by a person, not the assistant — she'll get back to you in the morning. I've saved your message.
A price question is answered straight from the catalogue. A refund question is held for the owner — every time.
Switching it off is one toggle
None of this is all-or-nothing. The assistant runs per channel, so you decide where it answers. Keep it on for Instagram and Facebook overnight, leave Telegram to yourself, or turn the whole thing off before a holiday. One toggle, no settings maze.
InstagramOn — answering 
FacebookOn — answering 
TelegramOff — goes to you Toggle any channel — the assistant only answers where you let it.
A quieter morning
The point was never to replace you. It's to make sure the night isn't wasted — that the easy questions are handled, the hard ones are waiting, and you start the day knowing exactly where things stand. Most mornings, that's a short list and a cup of coffee.
