How AI reads a chat and knows who's ready to buy
There's no crystal ball inside an AI like Aurora. Just a very patient reader that goes through every chat while you sleep, and reads the same tells you'd read yourself — if you had the hours. Here's exactly what it looks for, and how it decides who lands at the top of your morning.
First, let's kill the magic
When people hear "AI knows who's ready to buy," they picture a mind-reader. It isn't one. An AI doesn't predict the future. It reads. It goes through a conversation line by line and picks out the small signals a good seller already catches on instinct — "how much?", "is it still available?", the reply that came in ninety seconds, the one that never came at all.
The only real difference is stamina. It never gets tired, never skips a chat, never forgets the person who messaged you at 11pm on a Tuesday. You read three conversations well. It reads all four hundred, the same way, every single night.
It scores two things, not one
Most people lump "hot lead" into a single blob. A good AI splits it in two, because they answer different questions.
Intent is about what they want and how close they are to paying. "Do you ship to Recife?" is higher intent than "cute!" Temperature is about right now — is this conversation alive, or going cold? Someone can have sky-high intent and be ice cold because you left them on read for four days. Reading both at once is the whole trick.
The signals that mean "I'm ready"
These push a chat up the list. None of them are secret. You'd flag every one of them yourself.
Price questions — "how much?", "any discount for two?" Someone doing math is someone already imagining the purchase. Availability — "do you still have it?", "is the blue one in stock?" They've picked; they're just checking the coast is clear. Commitment words — "I'll take it", "send me the link", "can you deliver Friday?" That's a hand reaching for the wallet. And specifics — a size, a date, a delivery address, their name. Vague browsers don't hand over the details a real buyer does.
The signals that mean "slipping away"
Cooling is quieter, which is exactly why it costs you sales. The loudest cooling signal is time. A chat that was firing every two minutes and has gone silent for two hours is a different animal at hour forty-eight. The AI watches the clock on every open thread, so you don't have to.
It also reads the shape of the drop-off. Warm, warm, warm — then a one-word reply. Or the classic: you sent the price and they went quiet. That silence after a number is rarely a "no." It's usually someone who got distracted, or is waiting for a small nudge. That's the exact chat that becomes part of the Dark — a real sale lost not to rejection, but to a message nobody sent back.
How signals turn into a number
Here's the part that sounds complicated and isn't. The AI weighs each signal, and weighs the recent ones far more heavily. "I'll take it" from twenty minutes ago beats "love this!!" from three weeks back — every time. Freshness does a lot of the work.
Then it folds intent and temperature into one simple sort. High intent plus hot equals top of your list. High intent plus cooling equals your urgent saves — the ones worth a message before lunch. Low intent plus warm is polite, not a priority. It's not a mystical score out of 100 you have to trust blindly. It's "this person asked to buy and hasn't heard back — start here."
Why you get a ranked list, not a dashboard
A pile of unread chats is flat. Every bubble looks equally urgent, so none of them are. You answer top to bottom until your coffee's cold, and you never reach the person who said "send the link." A ranked list fixes the one thing that quietly costs you money: order.
That's the whole point of waking up to who's ready to close. Not more data — a starting point. You open the app and the first name is the person most likely to pay you today. You work down from there, and you stop before you burn the morning on browsers.
It's a copilot. You still close.
Now the honest part. This isn't perfect and shouldn't pretend to be. Sarcasm fools it — "great, another delay" can read as positive if you only count the word "great." The quiet buyer who says nothing for a week and then sends a full payment will blindside it. And "just looking" sometimes means "I'll buy in an hour."
So treat the score as a sharp assistant's opinion, not a verdict. Aurora reads the room and drafts the reply; you read it, tweak the tone, hit send — or don't. It takes the reading and sorting off your plate so your judgment lands where it's actually worth something: on the real human about to say yes.
And you can run a lo-fi version of this tonight, with no app at all. Open your DMs and re-sort them in your head by two questions — did they show they want it, and when did we last talk? Answer the person who's high-want and going cold first. That's the entire algorithm. The AI just does it for four hundred chats while you're asleep.