AI-generated posts and graphics for Instagram and Facebook without team involvement—automation for an educational brand

What problem did the educational brand face?

Julita Habet runs an educational brand for female and male network marketing leaders: online courses, webinars, and mentoring. Sales in this model rely on consistent engagement—a break in posting on Instagram and Facebook means a break in the sales funnel.

The problem wasn’t a lack of a plan. Julita teaches her audience how to plan content and works from a publishing calendar herself. The bottleneck was execution: between a ready-to-go topic and a published post lay the tasks of writing the content, preparing the graphics, and uploading it to two separate channels. The plan was created once, but it fell apart every day.

What did we build?

A system where planning remains in human hands, while production is automated. Here’s how it works:

1. Once a month, the client fills out the content calendar—just the topics, without writing the posts.
2. Automation pulls the topics from the calendar according to the scheduled publication date.
3. The GPT model uses these topics to write an engaging post in the brand’s tone of voice.
4. GPT Image generates graphics tailored to the post.
5. n8n orchestrates the entire workflow and forwards the finished post.
6. Postiz publishes it on Instagram and Facebook at the scheduled time.

Between steps one and six, the client doesn’t have to click a single time. The decision is made once—when creating the calendar. There’s no queue of posts waiting for approval, no “approve for publication” notifications, and no late-night review of drafts. Approving the topic means approving the publication.

This is a deliberate design choice, not a shortcut. Automation with a per-post approval step doesn’t solve the problem; it just shifts it: instead of writing twenty posts, you click “approve” twenty times—and you’re back to the daily interaction with the system you were trying to avoid. We’ve moved the checkpoint to where the actual decision is made anyway: to the calendar level.

The result is a change in how you work, not just a reduction in clicks: one planning session a month instead of figuring out every day what to post today. The calendar stops being a document you return to with a guilty conscience and becomes the gateway to a system that runs itself.

How does the automated system know what the brand’s voice sounds like?

The brand’s tone isn’t a side effect of the model here—it’s an integral part of the implementation. The automation system operates based on a brand voice profile built on how Julita actually writes to her audience, and it generates content tailored to that profile—not to the model’s default style.

This is where off-the-shelf post generators most often fall short: they produce text that is grammatically correct but completely indistinguishable from a thousand others. For a personal brand, this isn’t a minor detail—it’s the entire product.

Since there’s no post-by-post approval, the brand voice profile ceases to be a mere improvement and becomes a prerequisite for the entire model to function at all. No one will catch an off-target text before publication, so quality must be built into the generation process, not tacked on as a final check. That’s why fine-tuning the tone was a key part of the implementation here, rather than the last item on the list.

Why n8n and Postiz, rather than an off-the-shelf tool?

A ready-made scheduler handles just one step—publishing content you already have. It doesn’t pull topics from your calendar, write a post based on them in your brand’s tone, or create graphics. It takes care of the last 10% of the work, not the 90% that actually takes up your time. Tools that go further and generate content do so in the model’s default voice—which is exactly how everyone else does it.

n8n handles the orchestration of the entire workflow and can be deployed on your own infrastructure. Postiz takes care of the most tedious part—maintaining connections and tokens to Meta, as well as the publishing queue for Instagram and Facebook. GPT and GPT Image generate both content and graphics in a single run, instead of using two separate tools and manually piecing them together.

The most important difference, however, is legal, not technical: once the full payment is made, the economic copyrights to the automation solution transfer to the client. You’re not renting the process from us—you own it.

How did the implementation go?

Implementation took 7 business days from the start of work. After the system was handed over, a 7-day stabilization period began: we address all issues reported during this time at no additional cost, and once the corrected version is delivered, the period starts over.

This fully addresses the most common fear with such projects—“they’ll do something I don’t understand and then disappear.” The deadline, correction process, and support period were all agreed upon in writing before a single line of code was written.

Who is this model suitable for?

For companies and brands that publish regularly across two or more channels and are able to sit down once a month to plan out topics—but don’t have a team to turn those topics into finished posts. Online course creators, personal brands, offices, and agencies with a consistent publishing schedule.

It doesn’t work in two cases. First: every post is a separate creative campaign with its own concept—in that case, the automation system has nothing to work with. Second, and much more common: no one in the company is able to fill out the calendar. Automation enforces the plan, but it doesn’t create it. If the calendar remains empty, the system grinds to a halt—and that is, quite frankly, the biggest limitation of this model.