In short: the twelve minutes isn’t writing time, it’s decision time. How much to admit, whether to promise a date, whether “Sounds good!” reads as careless or warm. The typing takes twenty seconds; everything before that is a negotiation with someone who isn’t in the room. So the thing worth fixing was never how fast you type.
Watch someone reply to a slightly uncomfortable message and you’ll see the same ritual every time. They read it. They start typing. They delete. They retype. They read it again. They delete half. They stare at the screen, weighing a single word.
Twelve minutes later, they send two lines.
Two lines. Twelve minutes.
The typing took twenty seconds. So where did the other eleven minutes and forty seconds go?
Why does replying to one message take so long?
It went to a hundred silent negotiations:
- If I admit it isn’t done, do I look unreliable?
- If I explain why, does it sound like an excuse?
- Should I promise a time? What if I miss it again?
- Is “Sounds good!” too casual for my boss? Is “Received, working on it now” too stiff for a friend?
None of this is writing.
It’s a particular kind of load — the hesitation, the second-guessing, the fear of being read the wrong way. Almost nobody talks about it, because most of us quietly assume we’re just slow.
You’re not slow. You’re holding something the other person can’t see. The words are the cheap part.
What actually eats the twelve minutes?
Once you pull them apart, those minutes break down into four kinds of weight:
| What you’re doing | What it feels like | Why it takes so long |
|---|---|---|
| Choosing a position | Admitting, promising, declining | Each one changes what you owe afterwards |
| Calibrating tone | Too warm? Too cold? | There’s no correct answer, only a guess about how it will land |
| Re-reading your own draft | “Would I want to receive this?” | You move into the other person’s head and lose your own sentence |
| Checking one fact | “Did I actually verify that number?” | Because a fluent wrong reply costs far more than a slow right one |
Only the last one is about accuracy. The other three are about the relationship — and no amount of typing speed touches them.
That’s why shortcuts don’t fix it. A template that produces a reply you still have to second-guess has just moved the twelve minutes further down the road.
Why do I rewrite short messages over and over?
Because the cost isn’t composing sentences, it’s predicting how they’ll land. Short messages carry fewer cues, so every word holds more weight — which is why a two-line reply routinely takes longer than a page of writing.
Why this isn’t only about you
The obvious explanation for those twelve minutes is that you’re the kind of person who overthinks — you examine every sentence, you can’t tell when to stop.
Part of it isn’t you at all.
Which philosophy your tools picked
Start with the three most common ways people actually message each other:
| iMessage | Slack | SMS | |
|---|---|---|---|
| Tells the other person you read it | Yes — per contact, and never in groups | No | No |
| Unsend | 2 minutes | Yes | No |
| Drafts | Yes | Yes | No |
| Typing indicator | Yes | Yes | No |
The row that matters is the first one.
iMessage is the only one that shows it by default. Even there you can turn it off for one person at a time — useful, because switching it off globally also hides you from the people you actually want visible. It’s also off by default in group chats, whether or not anyone has it enabled.
Slack refuses. Not as a missing feature — as a position, and Slack has said why: telling the sender exactly when their message was opened creates pressure to reply the instant it’s seen. So the choice of whether to respond stays with the person receiving it.
SMS can’t. It’s a 1992 protocol. There is no read receipt, no unsend, no draft, no typing bubble. A text is a text.
Which is why the waiting feels different depending on which app you’re in. The ones that show you “read” teach you to expect being watched. The ones that don’t leave you with nothing — and nothing, it turns out, is its own kind of pressure: I can’t even tell whether they saw it, so what am I waiting for?
What actually makes “quick question?” different
English has its own warning shots. “Quick question?” “Just checking in?” “Can I ask you something?” — all of them the same move, buying a little room before something lands.
What makes it heavier is everything stacked around it:
- @here and @channel. In Slack, being @‘d in a channel means someone is waiting on a specific person, and that person is you. Neither works in a thread, so the pressure lands in the channel where everyone can see it.
- Time zones. Waiting on someone in London or New York turns a two-hour question into an overnight one. The clock is running whether or not anyone chose to start it.
- The always-on baseline. If you’ve ever answered something at 11pm, you’ve set the expectation that you might again.
None of that is your personality. It’s the shape of the tools, and the shape of the working day.
The apology you don’t owe
There is a whole genre of English-language email that opens with three lines of apology before it says anything. “Sorry for the slow reply.” “Apologies for the delay.” “So sorry to just be getting back to this.”
Adam Grant, the organisational psychologist at Wharton, has argued the phrase should be deleted from every email you send. He found 547 instances of it in his own sent mail in a single year. His case:
- It implies you have no work-life balance. Apologising for not replying instantly reinforces the idea that you were available around the clock. It says work is the dominant force in your life.
- You have nothing to apologise for. His line is the one worth keeping: if you didn’t commit to a deadline, you can’t be late.
- A good answer beats a fast one. He’d have us stop mistaking promptness for politeness.
If you must open with something, he suggests gratitude instead — “thanks for your patience” — which does the same social work without conceding a fault.
The apology isn’t the problem. The problem is that it turns your reply into a favour you’re granting, instead of an answer you’re giving.
There are two layers, and a tool only fixes one
Everything so far has been the first layer: organising the words, then sending them to the person.
Then you open an AI tool to help you answer one message, and you discover something: you have to write the message twice. Once for the person. Once for the tool.
“Help me reply — tell them I’ll have the numbers tomorrow.” And then you’re back to organising language: who they are, what the relationship is, what tone you want, which facts can’t be wrong.
The hard part didn’t go away. It moved.
And the second layer is harder than the first:
- The first layer is not knowing what to say
- The second layer is not knowing how to say “I don’t know what to say”
Almost nobody writes about the second layer. It’s probably the most real one now — the first is as old as letters, the second is something everyone using AI has hit.
A tool that only makes you type faster hasn’t touched it at all.
How much of the twelve minutes can actually go away
Honestly: not most of it.
Of the four kinds of weight, only the last one — checking a fact — can be handed to a tool. The other three are decisions. Picking a position, calibrating tone, re-reading your own draft: someone else doing those for you means someone else carrying them.
But three things can be moved:
- Say the decision out loud first. To yourself, once: what exactly am I agreeing to, and what am I declining? Then write. The words tend to follow on their own.
- Separate “must be accurate” from “just wording.” The first can’t be negotiated. The second can — and that’s the part a tool is actually for.
- Accept that “I don’t want to answer right now” is a legitimate answer. Sometimes those twelve minutes aren’t about finding words. They’re about not being ready. Parking it is more honest than forcing a reply.
What should a reply assistant actually lift?
Here is the thing we kept returning to while building BreatheReply: a tool that only helps you type faster solves the wrong problem. Autocomplete shortens the twenty seconds. It does nothing for the twelve minutes.
The twelve minutes is where help is needed — and the help isn’t writing for you. You don’t want a bot speaking for you to your team. The help is dissolving the hesitation:
- Read the scene first. Does this need warmth or firmness, brevity or detail? Settling that out loud, early, removes most of the steering later on.
- Offer directions, not one verdict. When a reply could reasonably go several ways, seeing them side by side turns a private agonising into a choice you make in one click.
- Ask when it doesn’t know. If a number or a deadline is yours alone, it should stop and ask rather than invent it — which is the reasoning behind why it asks instead of guessing.
You still decide. You still send. What gets lifted is everything sitting between you and the first word.
Doesn’t this make replies sound generic?
It should do the opposite. Nearly all the effort in a good reply goes into reading the situation correctly; once that’s settled, the wording stays yours. Edit it as much as you like — the draft is a starting point, not a verdict.
Is it wrong to paste a message into a chatbot for help?
It works, and plenty of people do it. But it sends the most private thing you wrote that day to someone else’s server, into logs you can’t see. If you use a cloud assistant for this, it’s worth knowing exactly where that text goes — we go through the difference here.
A few specific questions
Is twelve minutes really long?
Compared to what. Ten minutes on a reply that needed thought is a high ratio — but plenty of long gaps aren’t about wording. They’re about not having decided whether to answer. How much of your twelve minutes was really about choosing words, and how much about avoiding it?
Am I overthinking this?
People who don’t overthink still take twelve minutes. Their hesitation just doesn’t show. The difference is that afterwards they know what they were hesitating about, and you come away thinking I’m not the kind of person who texts back.
Will a reply written by AI sound like it isn’t mine?
It will, if the first draft is the last draft. Which is why a tool like this should hand you a starting point, not a finished sentence: it takes over the expensive part — reading the situation, offering a few directions — and leaves the wording for you to edit. Once you’ve edited it, the line is yours, because it passed through your judgement.
If we were to pick a way to measure BreatheReply, it wouldn’t be words per minute. It would be the gap between “I should reply to this” and “done.”
Not because speed is the goal. But because that gap is where the weight lives — and shrinking it is the reason the product exists.
Reply calmly. The words were never the hard part.
The twelve minutes don’t disappear because someone types faster.
And sometimes the twelve minutes means something else entirely: not “I can’t find the words,” but “I can’t face this yet.” That deserves respect too. Reply later lets you park a message with a preset delay or a custom time — so it can leave your attention without leaving it unresolved, and come back when the moment is better.
Join the waitlist — BreatheReply is built to shorten that specific gap: you select the message, and it gives you a direction to accept, warm up, or push down.