Home/Cases/Automating a small business's social media
6 hours a week down to 30 minutes How we automated a gym's social media without it losing the owner's voice.
A gym published one post a week, whenever the owner found a gap. We built a system that writes, designs and publishes five pieces a week in the business's own brand, with mandatory human approval before anything goes live. Tooling cost: $0 a month. The owner went from six hours a week to thirty minutes of review.
Nobody stops posting because they are lazy. They stop because each post costs three hours.
The actual loop was: come up with the idea, find a usable photo, write the copy, build the design, remember to post it when people are actually looking. Five steps, five different kinds of friction, all resting on one person, the same one working the front desk.
The output was not bad content. It was intermittent content: two solid weeks, then three weeks of silence, plus the permanent feeling of owing something. For a local business that is worse than not posting at all, because the algorithm learns the account is unreliable and so do the customers.
We measured it before touching anything, because without a starting number there is no case, only a story: six hours a week of thinking, writing and designing, for one post that went out whenever it went out.
The same week, before and after.
Left: the week done by hand. Right: the week as it runs now. Same business, same owner, same tooling budget.
Drag the handle · or use ←→ on your keyboard
Four pieces. Not one of them carries a monthly subscription.
This is not "an AI that posts". It is a chain with a human brake in the middle, and that brake is what lets the owner sleep at night.
-
1
The brand system
Before writing anything: palette, type, shapes and (the part that matters) the voice. Where the voice comes from: one session with the owner where we write down the ten questions their customers ask most, in the owner's own words. That document is what later keeps the posts from reading like a brochure.
Written once · corrected whenever something reads wrong -
2
The templates
Posts, carousels, stories and reels as templates for that business, filled with data and rendered to image or video at exact Instagram dimensions. Nobody nudges a text box again: the design is settled, and every piece comes out equally clean.
HTML + automated render to PNG and MP4 -
3
The queue and the checks
A board showing what is coming: what publishes, when, and what state it is in: idea, written, rendered, approved. An automated checker validates the brand rules before a human even looks. And here is the brake: nothing moves forward until a person approves it.
Owner approves from their phone · about 10 minutes -
4
Publishing
Approved pieces publish themselves at the chosen time through Instagram's official API, the same one agencies use. If anything breaks there is a plan B: the content pack is already built and goes up manually in two minutes. The system lives in the business's accounts, not ours.
Official API · $0/mo · no lock-in
What changed, measured.
Running in production for months. Rounded figures from the real business, not a projection.
| What we measured | Before (by hand) | After (with the system) |
|---|---|---|
| Owner time per week | ≈ 6 h | ≈ 30 min |
| Pieces published per week | 1, irregular | 5, every week |
| Building one post | ≈ 3 h | ≈ 10 min of review |
| Tooling cost | $0 (but 6 h of someone) | $0/month |
| Human approval before publishing | Yes | Yes, still mandatory |
Our own case, generalized for confidentiality. Rounded figures from a system in production, not projections. The system does not raise sales on its own: it frees the time and holds the consistency, which is what moves the needle afterwards.
We built it twice. The second time took a fraction as long.
After the gym we ran the same engine for ourselves. That confirmed what we suspected: the brand changes, the machinery does not. Which is why implementation N+1 always costs less than N, and why this ships in weeks, not months.
Identical across businesses
- The image and video render engine
- The editorial queue and its states
- The automated brand-rule checker
- Scheduled publishing and access handling
- The golden rule: human approval before every post
Rewritten per industry
- Your customers' real questions and pain points
- The look of the templates
- The examples and cases the content tells
- The hours your audience is actually watching
What this system does not do
It does not answer DMs on its own (that is a separate module). It does not invent results. It does not publish without your sign-off. And it does not lock you in: it is documented and lives in your accounts, so any technical person can maintain it.
Questions we get every time
Won't AI-written posts sound like a robot?
The voice comes from a playbook written with you: your words, your customers' actual questions, your examples. And nothing publishes without your approval. When a draft reads wrong, we fix the playbook, not just that post, so the mistake does not repeat.
What does it cost to run each month?
Zero in tooling. Everything runs on free tiers: Instagram's official API, a free automation tier, and a temporary image host. The only cost is the initial build project.
Does it publish on its own, with no one looking?
No, and that is deliberate. The system prepares everything and waits. You review from your phone, approve or correct, and only then does it go out at the scheduled time. About 10 minutes a week.
Does this lock me into working with you?
No. The system lives in your accounts and is documented. Any technical person can maintain it. We would rather say that upfront than have you find out later.
How long does it take to build?
Four phases across roughly three to four weeks: brand system, templates, engine and queue, then scheduled publishing. Weeks, not months, because only the brand layer is new each time. The engine is already built.
See the real thing, not a screenshot.
Answer seven questions first (two minutes, no contact details until the end). On the call that follows we demo this system live: the actual queue, the actual approval step, the actual publishing. Then we tell you honestly whether your business needs it.
Start the free assessment→ Or read the other cases first.