Drop an image,
get the post written.
A client of mine runs a social media agency, and her team was drowning. The slow, boring part of the work was taking each client’s images, writing the right caption, finding the hashtags and the tags, then scheduling it all. It ate whole days and whole people, so they could not take on more clients. So I built a machine for it. The client drops their images into a folder, and a model I trained on that client’s own past posts reads each image, works out what kind of post it is, applies the right framework, and writes the caption, hashtags, and tags in the client’s exact voice. A batch that took a full day is done in minutes. Everything then lands in an approval queue, where the client checks it and marks the ones they like. The moment a post is approved, a second bot sends it straight to the scheduler. I deploy this for clients, and it lets one person run an agency that used to need a room full of them.
Writing captions for a room full of clients is a full day of boring work. The machine does it in minutes.
Every social media agency runs on the same slow grind. The client sends the images, and someone has to write the caption, dig up the right hashtags, add the tags, and schedule the post. It is not hard work, but it is endless, and it swallows whole days and whole people. That is why so many agencies cannot grow. They are too busy writing captions to take on the next client. So I stopped the grind at the source. The client drops images into a folder, and a machine trained on their own past posts writes each post in their voice, sends it to an approval queue, and once approved schedules it. Here is who it is for, what goes wrong without it, how it works, and what you get back.
Who social media posts automation is for
Anyone who produces social posts for other people. Social media agencies, freelancers who manage accounts, marketing teams that juggle several brands. It fits best where the same slow job repeats across many clients, because that is exactly the work the machine takes off your plate.
What goes wrong without social media posts automation
Without it, caption writing eats your day and caps your growth. You cannot take on more clients because your people are buried in manual work. You are not short on demand. You are short on the hours it takes to write and schedule every post by hand, one at a time.
How the social media posts automation automation works
The client drops images in a folder. A model trained on their past posts reads each image, picks the right framework, and writes the caption, hashtags, and tags in their voice. It routes each post to an approval queue, and once approved a second bot sends it to the scheduler. You write nothing by hand.
What social media posts automation gives back each month
A full day of writing turns into a few minutes of checking. Your people are freed to take on more clients instead of grinding through captions. And because agency work is recurring, every client you can now serve at the same cost adds revenue that comes back month after month.
She had the clients lined up. She just could not write the posts fast enough to take them.
A prospect came to me with a real problem. She runs a social media agency, and she has to post content for her clients. The slow, extremely time-consuming part of the job was receiving the images from the clients, then writing the right caption, finding the right hashtags and tags, and scheduling it all on the client’s behalf. It took so much time and so many people that they could not accept more clients. They were too busy and too stressed to step back and grow.
So I built a machine that does in minutes what used to take them a whole day. The client gets a folder where they drop their content. Before anything runs, I train a model on all of that client’s past posts, so it understands how they write, why they write that way, the kinds of posts they make, and the frameworks behind each type. The model learns the client’s voice from the client’s own work.
Once an image lands in the folder, the machine starts running. It figures out what kind of post this is, chooses the framework that fits, and writes the caption, the hashtags, and the tags so they are relevant to the image and congruent with how the client normally writes. It just sounds like them. Drop in a batch and it works through them together, complete posts done in the time it takes to make a coffee.
It does not stop there. You do not want to leave everything to the machine, so each finished post goes into a validation queue. The client checks it, and if they like it, a simple drop-down lets them mark that post as ready to publish. The moment a post is approved, a second bot grabs the image, the caption, the tags, and the hashtags, and sends the whole thing to the scheduler, where it gets slotted in on its own.
The numbers here are kept deliberately modest. Say the machine turns out a few posts a minute instead of the flood it can actually manage. A day of caption writing shrinks to minutes of checking, so one person can serve the clients that used to need a small team. The agency takes on more clients while the costs stay flat, and because the work is recurring, that extra revenue comes back every month.
Proof point: I have documented how I run my business on autopilot and the income streams behind it on YouTube, in my 28 income streams breakdown and a real look at the daily work in a day in my life, so the conservative numbers on this page are checkable.
Three moves that turn a folder of images into scheduled posts in the client’s own voice
What made this work was breaking the job into three clean moves and letting the machine own the boring middle. Most agencies do all three by hand, over and over, for every client, and it never ends. Here the client drops the images, the machine writes each post in their voice, and once a human approves it, another bot schedules it. The person only does the two things that need a person, sending the images and saying yes. Everything in between runs on its own.
The client drops the images and the machine reads them
The loop starts the moment a client drops an image into their folder. There is no brief to write, no instructions to type. The machine picks it up on its own and reads it, then works out what kind of post it is. A product shot, a behind-the-scenes photo, a quote card, each gets recognised for what it is. That matters because the type of post decides the framework, and the framework decides how the caption should be written. The client did nothing but drop a file, and the machine already knows how to handle it.
It writes the whole post in the client’s exact voice
This is where the training pays off. Before it ever writes a word, the model has studied all of that client’s past posts, so it knows their style, their tone, and the frameworks they lean on. For the image in front of it, it applies the right framework and writes the caption, the hashtags, and the tags so they fit the image and sound like the client. Not generic, not robotic, but congruent with how that client actually writes. Drop in ten and it turns them around in about a minute, each one reading like the client wrote it themselves.
A human approves, then a bot sends it to the scheduler
You never let the machine publish blind. Every finished post lands in a validation queue where the client reviews it and, with a simple drop-down, marks the ones they like as ready. That is the only judgment call a person makes. The instant a post is approved, a second bot grabs the image, caption, hashtags, and tags and sends the whole thing to the scheduler, where it gets slotted in on its own. Control stays with the human, the grind goes to the machine, and nothing goes out that the client did not approve.
Once those three moves are in place, the manual grind stops. The client drops the images. The machine writes each post in their voice and files it for review. The client approves the good ones, and a bot schedules them. What took a full day and a small team now takes minutes and one person, and the quality still sounds like the client every time.
Before the system
- Writing every caption, hashtag, and tag by hand, one at a time
- Whole days and whole people lost to slow, boring work
- Turning clients away because the team was already buried
- Captions that drift off the client’s real voice under pressure
- Growth capped by how fast people can type, not by demand
After the system
- The client drops images and the posts write themselves
- A day of work done in a few minutes of checking
- One person serving the clients that used to need a team
- Every caption in the client’s own voice, learned from their posts
- Approved posts sent straight to the scheduler on their own
Prompt 1: learn a client’s voice from their past posts
The machine only sounds like the client because it studied the client first. Before you automate anything, you map the voice. Use this prompt to pull the patterns out of a client’s past posts so every caption matches how they really write.
Client voice mapper
Act as a brand voice analyst. I run social accounts for clients and I want to capture one client's writing voice so every future caption sounds like them, not like a generic bot. About this client: [paste 15 to 20 of their past captions and describe their business]. Map their voice: the tone and personality, the sentence length and rhythm, the words and phrases they lean on, the kinds of posts they make, and the framework behind each type. Give me a short voice guide I can reuse for every new post, and one line on why each pattern matters.
The output is a reusable voice guide for that client. Feed it to the machine and every caption it writes lands in their real tone, not a stock one.
Prompt 2: detect the post type and pick the right framework
A quote card and a product shot are not written the same way. The machine has to recognise the type before it writes, so the framework fits. Use this prompt to build the rules that sort each image into the right kind of post.
Post-type framework picker
Act as a social content strategist. When a new image comes in, I want to work out what kind of post it is and apply the right caption framework, so the writing always fits the content. About my clients' content: [list the types of posts they publish, for example product shots, behind the scenes, quotes, tips, announcements]. Build the rules: for each post type, how to recognise it from the image, the caption framework that fits, the hashtag and tag approach, and one thing to avoid. Keep it simple enough to run on every image with no guessing.
The output is a clear map from image to framework. Now every post the machine writes starts from the right structure, so the caption always fits what is in the picture.
The 3-minute overview of how this works
Before the build steps, watch this short overview. It’s the exact video from the Automations Made Easy page, and it walks through the mechanics behind machines like this one. 1,000+ students have used these mechanics to save two hours a day, with zero coding.
Want to learn the mechanics behind content production automations like this?
I teach the same mechanics that power this social posts machine inside Automations Made Easy. 1,000+ students have used these mechanics to save two hours a day and build little machines that quietly handle the boring work. No coding required.
Week one: your first client’s folder live, the first posts written in their voice.
Most people put off building a content machine because it sounds complicated, so they keep writing every caption by hand and stay too busy to grow. That is exactly why this machine matters. A content machine is meant to start with one client and grow, not arrive perfect. In week one you set up one folder, train the model on that client’s past posts, and watch the first captions come out in their voice, so you know it works before you roll it out to everyone.
Week one looks like this. On Monday you gather one client’s past posts and use them to teach the machine their voice and their post types. Then you set up their folder and drop in a few test images to see what comes back. By midweek you connect the output to an approval queue, so finished posts land somewhere the client can review them. By the end of the week you wire the approved posts through to the scheduler, and your first real batch is written, checked, and scheduled without a caption typed by hand.
The point of the first week is not the number of posts produced. The point is to confirm the machine reads a dropped image, writes the caption, hashtags, and tags in the client’s voice, files it for approval, and schedules what gets approved. Once that loop is locked for one client, you copy it for the next, and the day of manual writing turns into minutes of checking, every time.
From there the maths is simple and conservative. Say the machine writes a few posts a minute, and a batch that took most of a day is done before your coffee cools. That freed time is time your people spend on new clients instead of captions. Add even a couple of extra clients a month at the same cost, and because agency work is recurring, that revenue compounds over the year and keeps coming back.
All of this runs while you work, sleep, or pitch the next client. The client drops the images, the machine writes the posts in their voice, files them for approval, and schedules what gets the nod. No more full days lost to captions, no more turning clients away because the team is buried. The slow part of the agency stops being a ceiling and becomes something one person can quietly carry for many clients at once.
Prompt 3: write the caption, hashtags, and tags for one image
This is the heart of the machine, the prompt that turns one image into a finished post in the client’s voice. Use it with the voice guide and framework rules you built above, so every caption fits both the client and the content.
Post writer in the client’s voice
Act as a social media copywriter writing in a specific client's voice. Using the voice guide and the framework for this post type, write a complete post for the image I describe. Client voice guide: [paste the voice guide from prompt 1]. Post type and framework: [paste the matching framework from prompt 2]. The image: [describe what is in the picture]. Write: a caption in the client's exact tone, a set of relevant hashtags, and the right tags, all congruent with how this client normally writes. Keep it ready to post, no placeholders.
The output is a complete post that sounds like the client, not like a machine. Run it across a folder of images and you have a batch done in minutes.
Prompt 4: set up the approval queue and the hand-off to the scheduler
A finished post still needs a human yes before it goes out. This prompt maps the validation queue and the bot that sends approved posts to the scheduler, so control stays with you while the grind stays with the machine.
Approval and scheduling planner
Act as a marketing automation advisor. I want finished posts to land in an approval queue where the client marks the good ones, then have a bot send only the approved ones to my scheduler. About my setup: [name the tool where posts are reviewed and the scheduler you publish through]. Plan it: how the client reviews and approves each post with a simple status change, what the bot grabs when a post is approved, the image, caption, hashtags, and tags, how it sends that to the scheduler, and how to make sure nothing unapproved ever goes out. One line on why each step matters.
The output is a clean path from approval to scheduled post. The human keeps the final say, and the machine handles everything after the yes.
How to build social media posts automation, step by step
Give the client a folder to drop their images in
Start where the work begins. The client already has the images, they just need a simple place to put them. So you give each client a folder that feeds the machine. No brief, no instructions, no email back and forth. They drop the picture in and walk away. This is what makes it painless for them and for you. The client does the one thing only they can do, hand over the content, and from that moment the machine takes over. A dropped file is the trigger for everything that follows.
Train the model on the client’s own past posts
This is the step that makes everything sound like the client. Before the machine writes a single caption, you feed it all of that client’s past posts. From those, it learns how they write, why they write that way, the kinds of posts they publish, and the framework behind each type. It builds a model of the client’s whole style. This is done once, up front, per client. After that, every caption the machine writes draws on that training, so it reads like the client, not like a stock tool that could be writing for anyone.
Let it detect the post type and pick the framework
Now the machine runs. When an image lands in the folder, it first works out what kind of post it is. Is this a product shot, a quote, a behind-the-scenes photo, an announcement. It analyses the image and decides. That matters because the type of post decides which framework to use, and the framework decides how the caption should be built. A tip post is written differently from a product launch. By naming the type first, the machine makes sure the writing that follows fits the content in the picture, every single time, with no one sorting posts by hand.
Write the caption, hashtags, and tags in their voice
Here the whole post gets written. With the framework chosen and the client’s voice already learned, the machine produces the caption, the hashtags, and the tags for that image. Everything is relevant to the picture and congruent with how the client normally writes, so it sounds like them, not like a bot. Drop in ten images and it turns them all around in about a minute, each one a complete post. This is the step that replaces most of a working day, and it does it without ever getting tired, bored, or off-brand.
Send each post to a validation queue for approval
You do not want the machine publishing on its own, so every finished post goes into a validation queue first. The client opens the queue, reads the caption, and decides. If they like it, a simple drop-down lets them change the status to ready to publish. If they do not, it stays put. This is the one place a human is in charge, and it is the right place. The client keeps full say over what goes out under their name, while the writing that used to eat their day is already done for them.
Let a second bot send approved posts to the scheduler
Final piece. The moment a post is marked ready, a second bot picks it up on its own. It grabs the image, the caption, the hashtags, and the tags, and sends the whole thing to the scheduler, where it gets slotted in at the right time. No copying, no pasting, no logging into another tool by hand. The human said yes, and the machine did the rest. This is what turns a folder of dropped images into a full calendar of scheduled posts, all in the client’s voice, with almost none of your time spent.
Drop the image in
The client drops an image into their folder, and that dropped file starts the machine.
The machine writes it
Trained on the client’s past posts, it detects the type and writes the whole post in their voice.
The client approves
Each post lands in a validation queue where the client marks the good ones ready to publish.
A bot schedules it
The moment a post is approved, a second bot sends the whole thing straight to the scheduler.
Build this social posts machine inside the same playbook 1,000+ students use
Automations Made Easy teaches the mechanics behind content production machines like this one. Step by step, no code, plain English. Save two hours a day and own little machines that write and schedule the boring work for you, while you spend your time winning new clients.
The six months after I switched it on
Here is the shape of the first six months after I turned this machine on for the agency. The line tracks the number of client posts it wrote and scheduled each month, the boring work that used to eat whole days by hand, now handled by the machine while the team took on more clients.
Monthly client posts written and scheduled by the machine
Three things matter on this chart. The count climbs steadily as the agency adds clients without adding staff, not in a spike. The work all happens at the same cost, since the machine writes and schedules while one person reviews. And every post on it went out in the client’s own voice, checked and approved before it ever published.
What other students built with social media posts automation
I teach the simple skills behind machines like this in Automations Made Easy. Students who built their own version sent back what changed in their first month.
“I run a small agency and I used to spend my mornings writing captions. Now my clients drop images and the posts come out sounding like them. I check them and they schedule themselves. I took on three new clients without hiring anyone.”
“The part that sold me was the voice. It learned how my client writes from her old posts, and now the captions read exactly like her. She approves them in the queue and they go straight to the scheduler. She thinks I hired a whole team.”
“A full day of writing and scheduling turned into about twenty minutes of checking. I just open the queue, tick the good ones, and a bot sends them off. The rest of my day is finally free to actually sell.”
“Knowing week one was just one client’s folder kept it simple. By month two I had every client on it, the posts writing in each of their voices, and me reviewing instead of typing. My costs never moved.”
What’s inside Automations Made Easy
AME isn’t a library of pre-built automations. Every business is slightly different. What’s reusable across all of them is the underlying mechanics: how to set up little machines that listen, write, and follow up while you sleep, and how to wire the pieces together without writing code.
The program walks you through six modules: The Right Tools (the cost-effective, no-code stack I actually use), Task Selection Mastery (which automations are worth building first), Design Secrets (mapping an automation before you build it), Zero to Hero (complete beginner to confident automator), Real-World Application (we build a full automation together, end to end), and Monetization Mastery (turn the skill into a side-business).
It also includes done-for-you templates you import in two clicks, over-the-shoulder training videos, and the same playbook 1,000+ students have used to save two hours a day. No coding required. If you can copy and paste, you can build this.
The social media posts machine: common questions
Pulled from what readers and Automations Made Easy students ask most.
Do I need to be a developer to set this up?
No. You point a folder at the machine, feed it a client’s past posts so it learns their voice, and wire the finished posts to an approval queue and then a scheduler. The skills you need are knowing your client’s style and checking the output, which is exactly what Automations Made Easy teaches. The writing, the hashtags, and the sending are handled for you.
How does the machine sound like my client and not like a robot?
It learns from the client’s own past posts before it writes a single caption. It studies how they write, the kinds of posts they make, and the frameworks behind each type, then it matches that voice. When a new image comes in, it works out which type of post it is, applies the right framework, and writes in the client’s exact tone, so the caption reads like the client wrote it.
Does the automation post without me checking it first?
No, and that is on purpose. Every finished post lands in an approval queue where you or the client reviews it. Nothing goes out until someone marks it approved from a simple drop-down. Only then does the second bot pick it up and send it to the scheduler. You keep full control over what gets published, with none of the manual writing.
How many posts can it really turn around at once?
Drop a batch of images into the folder and it works through them together. To keep the numbers modest, think a few posts a minute rather than a flood, complete with caption, hashtags, and tags in the client’s voice. Work that used to take a person most of a day is done in the time it takes to make a coffee, and it never gets tired or bored.
How does this help me take on more clients?
The bottleneck in an agency is people-hours spent writing and scheduling. When that day of work shrinks to minutes, one person can serve the clients that used to need a small team. Your costs stay flat while you add clients, and because agency work is recurring, that extra revenue arrives every month for the same effort.
More Money Makers like this one
Built from the same handful of mechanics. Each one captures, converts, or earns money that would have walked out the door.
Build this social posts machine yourself, or learn the mechanics inside Automations Made Easy.
If you want to learn the mechanics behind content production automations like this and build your own at home, Automations Made Easy is the playbook. Step by step, no code, plain English. If you want to talk through how to train the model on a client’s voice, set up the folder and approval queue, and wire it to your scheduler, I take a small number of consulting clients each month.
€497 one-time · Lifetime access · 1,000+ students
