It's 11:47 at night. A woman is lying in bed, scrolling, and she stops on a reel: a linen dress in sage green, exactly the thing she's been hunting for since spring. She taps the DM button and types: "Hi, is the green one available in M? How much with shipping?"
The seller is asleep. Of course she is. She's been answering messages since 8 a.m.
At 7:50 the next morning the seller replies: "Hi dear, yes we have M! 🙏"
The customer has already bought a different dress. From a page that answered at 11:49.
Nobody did anything wrong here, which is what makes it so annoying. So let me show you what this quietly costs, why the damage happens in minutes rather than hours, and what a 24/7 AI customer response setup should and shouldn't do while you sleep.
Your shop keeps two sets of hours. You're awake for one.
When I look at seller inboxes hour by hour, the pattern repeats so often it's almost boring. There's a daytime bump around lunch and early afternoon. Then a second peak starts after dinner and runs well past 11 p.m. For plenty of pages, the evening peak is the bigger one.
It makes sense. Evenings are when people finally sit down with their phones. They aren't sneaking a look at work; they're on the couch, with time to decide. That's the highest-intent window of the day, and it lines up almost perfectly with the hours a one-person shop is offline. I broke the full curve down in the piece on the busiest DM hours and the peak sellers sleep through.
Then there's the other gap: the rush. You post a reel at 6 p.m., it catches, and 40 messages land in half an hour. You're awake, you're trying, but you can type maybe one decent reply a minute. Message number 38 waits 40 minutes.
For that customer, you were closed too.
So "24/7" is a slightly misleading label. The real problem isn't only the night. It's every minute where the messages outnumber your thumbs.
What an unanswered night actually costs
Let's run the numbers on an ordinary page. Not a giant brand. A small shop doing fine.
- You get 60 DMs a day.
- About a third arrive after 10 p.m. or during a rush you can't keep up with. Call it 20 messages.
- Most first messages aren't buyers. In the conversations we've analyzed, roughly 15% of first-contact DMs come from people genuinely ready to buy. That's 3 real buyers in the late pile.
- Say half of them won't wait until morning and buy somewhere else. That's 1.5 lost orders a night.
At a $40 average order, that's $60 a night. Around $1,800 a month, walking out of a shop that had the product, the price and the customer's full attention. You paid for the reel. You earned the follower. Then you lost the sale in the last three feet.
Your numbers will be different. Plug them in anyway. Most sellers I talk to have never done this sum, and the ones who do usually go quiet for a second.
And the loss isn't only the order. A customer who waited nine hours for "hi dear, yes 🙏" learns something about your page: they're slow. Next time your reel shows up in her feed, she remembers.
The first five minutes do most of the work
Here's the part people underestimate. The drop-off isn't gradual. It's a cliff.
The best-known research on this is an older Harvard Business Review study of online leads. Companies that responded within an hour were nearly seven times more likely to qualify the lead than those that waited even one hour longer, and more than 60 times more likely than companies that took a day or more. That study was about B2B sales teams, and a DM shopper is less patient than a procurement manager, not more.
What I see in chat fits that. A reply in the first few minutes catches the customer while the app is still open, the product is still on screen, and the want is still warm. After that she closes the app. When she comes back, she's comparing three pages instead of looking at one. If you want the full curve, I wrote about the five-minute window and how response time moves sales conversion.
Which means a nine-hour reply and a 90-minute reply are closer in outcome than you'd hope. The real line runs between "answered while they were still here" and everything else.
A human can't hold that line 24 hours a day. Not with two peaks, not with a reel that takes off at dinner time, and not with a life.
What always-on AI should actually do at 2 a.m.
This is where I'll take a side. A lot of what gets sold as "24/7 auto-reply" is a nicer-looking out-of-office message. "Thanks for reaching out! We'll get back to you during business hours."
That's not a response. That's a receipt for a lost sale.
A real 24/7 AI customer response has to do the job a good salesperson does in those first minutes:
- Answer the actual question. Is the green one in M? Yes or no, from your real stock and catalog, not a guess.
- Give price and shipping in the same reply, because that's always the next question anyway.
- Move toward the order. Ask for the size, the address, the payment step, whatever your flow is. A good reply ends with a next step, not a full stop.
- Sound like your page. If you write warm and casual, the assistant writes warm and casual. The customer shouldn't feel a shift change.
- Know when to stop. Anything outside what it knows gets handed to you, with context, instead of invented.
That last one matters more than the other four combined. The fastest wrong answer is worse than a slow right one, because a customer who gets a wrong price at midnight doesn't come back to argue. She just doesn't come back.
What it should leave for you in the morning
The goal isn't to remove you from your own shop. It's to make sure that when you wake up, the easy 80% is already done and the hard 20% is waiting with notes attached.
Things I'd never want an assistant improvising at night:
- Refunds, and complaints about a damaged order.
- Custom deals: bulk discounts, "can you do it cheaper if I buy three".
- Any conversation where the customer is already angry.
- Questions your catalog genuinely doesn't answer.
For those, the right move is a short, honest reply ("I've passed this to the team, you'll hear back first thing in the morning") and a clean handoff. I go through where to draw that line in the escalation rules for when AI should hand off to a human.
That customer still waits. But she waits knowing someone has it, which is a completely different experience from waiting in silence.
Setting it up this week
You don't need a big project. You need one honest afternoon.
- Pull your last 100 DMs and sort them by the hour they arrived. Count how many landed while you were asleep or buried.
- List the 15 to 20 questions that make up most of those messages. Price, sizes, stock, delivery time, payment, returns.
- Clean up your catalog. Current prices, what's in stock, what isn't. An assistant is only as right as what you give it.
- Write your handoff rules in plain words. What it should never answer on its own.
- Read the night conversations every morning for the first two weeks. Fix mistakes at the source, not by apologizing to one customer at a time.
After those two weeks, most sellers stop reading every chat and just skim the handoffs. That's roughly when it starts giving you your evenings back.
Back to the green dress
Rerun that night with an assistant switched on. 11:47, the question arrives. 11:47, the answer goes out: yes, M is in stock, here's the price, delivery takes three days, shall I reserve one for you? She sends her address from bed. At 7:50 you wake up to an order instead of a missed chance.
That's the whole idea. Not a robot replacing you. Your shop, answering in your voice, at the hour the customer actually decided to buy.
If you want to see that on your own page, Vardast connects to Instagram, WhatsApp, Telegram and your website, answers customers around the clock from your own product information, and hands the tricky conversations to you. Start with one channel and read the night chats for a week. The numbers usually make the decision for you.
Your customers aren't going to start shopping during your office hours. So the shop has to be open during theirs.