Everyone has received the bad version: a wall of text with a link, sent by an account that clearly didn't read anything, three seconds after you commented. It converts nobody and it costs the sender something worse than a sale — the follower's benefit of the doubt.
The good version is almost embarrassingly simple: short messages, in your voice, that deliver before they ask, and hand off to a real human the moment the conversation earns one. Here is how to write it.
Give every automation exactly one job
The automations that feel robotic are usually trying to do three things at once: deliver the guide, pitch the offer, and collect an email, all in message one. Each extra job makes every job weaker.
Decide the single job before you write a word. “Deliver the reset guide to people who comment RESET” is a job. “Nurture my audience” is a mood. When the job is one sentence, the messages write themselves — and when a rule needs a second job, that's not an edit, that's a second automation.
One trigger, one intent.
The person who comments GUIDE wants the guide. Meet the intent they expressed, not the one you wish they had.
One next step per message.
A message with two asks gets neither. Deliver, then offer one clear thing to tap.
One ending.
Every path through the automation should land somewhere deliberate — a link, an answer, or a graceful stop. No dead-end loops.
Write it the way you'd type it
Open your actual sent DMs and look at how you write to a friend: short lines, first names, the occasional emoji, no “Greetings!” The automation is borrowing your voice — it should sound like the best version of how you already talk, not like a brand found a thesaurus.
The test is brutal and useful: read the message out loud. If you wouldn't say the sentence to a person standing in front of you, cut it. “Here's the guide! The breathing one is my favourite — start there” survives the test. “We appreciate your engagement with our content” does not.
People don't mind that the reply was instant. They mind when it pretends to be something it isn't — or sounds like nobody at all.
Treat every tap as consent, not a conversion
A quick-reply chip — “Send it over ⚡” — looks like a UI detail. It's actually the ethics of the whole system: the person chooses, with one tap, whether the conversation continues. That tap is what opens Instagram's 24-hour window for links and follow-ups, and it doubles as the person saying yes.
Design around that consent. The chip's label should say what happens next in the person's words, not yours. After the tap, deliver immediately — the fastest way to lose the trust a tap grants is to answer it with a qualifying quiz.
Label the chip with the outcome.
“Send the guide” beats “Learn more.” The person should know exactly what the tap buys.
Never punish the tap.
Tap → thing. Any questions you genuinely need come after delivery, and one is usually the limit.
Silence is an answer.
No tap means no. The follow-up for silence is nothing — the next post will offer a fresh chance.
Meet people on every surface they already use
Comments are the loudest trigger, but they're not the only one. Story replies are higher intent than comments — the person typed to you, in a private context, about something you posted an hour ago. A story-reply automation that answers with the promised resource feels like magic precisely because the moment is so fresh.
DM keywords close the loop from everywhere else. “DM me the word PLAN” works in a caption, a live, a talking-head video, or a podcast — any surface where a link can't live. Same rule, same delivery, same voice; the trigger just meets people where they already are.
Story replies
Answer the reply with the thing the story teased. The context is minutes old — speed is the whole gift.
DM keywords
One word, sendable from anywhere, that starts the flow without needing a post to comment on.
Same voice everywhere
The person doesn't experience your triggers separately. Every surface should sound like the same human.
Know the moment a human takes over
Automation is brilliant at the predictable ninety percent: deliver, confirm, route. It is terrible at the ten percent that matters most — the nuanced question, the frustrated customer, the person who typed three paragraphs about their situation. Those messages are not a failure of the automation; they are its best output. A real person raised a real hand.
Build the handoff in from day one: real questions land in your inbox and get you, not a scheduled message three. And when someone asks to stop, stop — instantly, permanently, without a “before you go.” The person who left cleanly comes back; the person who had to fight their way out doesn't.
Measure conversations, not sends
“10,000 DMs sent” is a cost, not a result. The numbers that describe reality: how many people tapped or replied, how many reached the thing you promised, and how many wrote back something a bot couldn't have prompted — a thank-you, a question, a story.
Improve one message at a time, oldest bottleneck first. High deliveries with low taps is a first-message problem. High taps with low follow-through is a destination problem. And a rising number of genuine replies is the metric nobody puts on a dashboard — the automation is starting conversations worth having.
The bar isn't “sounds human.” It's “sounds like you.”
Deliver before you ask. Let the tap be the yes. Hand the hard messages to a person. Automation done this way doesn't replace the relationship — it's the part of the relationship that never sleeps.
Put it into practice
