Skip to content
Aldridge Dagos Get in touch

CASE 08

Social Media Content Engine

Running social used to mean writing every caption, making every image, and chasing approvals. Now the operator plans a post and approves it. The studio writes the words, generates the images, and publishes on its own.

Role
Design, build, automate
Client
Agency and in-house teams
Year
2026
Stack
Web app · AI generation · Automation
Status
Live in production

01 The problem

Running social is a grind of small jobs that repeat every day. Someone writes the caption. Someone makes the image. Someone posts it to the right account at the right time. Then the approvals start, a draft goes out by email, a reply comes back with a change, a new version goes out, and the thread runs for days before anything is live. A full team spends most of its hours on craft work that looks the same every morning, and the client and the team rarely see the same picture at the same time. The cost is real and it scales with output. More posts means more writers, more designers, and more rounds of approval, so volume always runs into the size of the team.

02 The system

A studio that does the making and a board that does the approving. The operator plans a post in one row, the topic, the networks, and the look, and presses generate. The studio writes the caption and creates the images for it, grounded in the brand kit and the recent posts, so every draft sounds like the brand and matches what went out before. The draft lands on a review board. Approve it and it schedules itself. Send it back with a note and it rewrites itself from that note, no editor, no new email thread. The two things a person does are plan a post and approve a post. Everything between those two clicks, the writing, the images, the checking, the scheduling, and the publishing, runs on its own. The daily grind of a content team collapses into two human actions, and output stops being capped by how many people you can put on captions and design.

03 How it holds up

The generator checks its own work before a person ever sees it. A guard reads every line against the brief and rejects any invented stat, price, or quote, then regenerates until the draft is clean. So the post you approve is grounded in real facts, not something a machine made up. Revision is a real loop, not a handoff. The note a reviewer leaves becomes the instruction the post rewrites itself from, and the card moves back onto the board ready for another look. Approved posts publish to Facebook, Instagram, and LinkedIn through a worker that claims each job once and never double-posts, so nothing goes out twice and nothing falls through. Spending is capped at four levels with a live meter and a kill switch, so the running cost stays visible and under control. The platform is multi-tenant and runs as a separate branded instance per client, so an agency can put every account on it without the work bleeding across clients.

04 The result

A team plans and approves. The portal writes, designs, checks, schedules, and publishes. The work that used to fill a day for a whole content team now runs on its own, so the same people ship far more posts without adding writers or designers. Nothing goes out that a person did not approve, so you get the volume of an automated pipeline with the control of a human sign-off on every post.

· How it works

  1. 01

    Plan a post

    The operator fills one row in the studio, the topic, the platforms, the look, and presses generate.

  2. 02

    The studio creates it

    It writes the caption and generates the images for the post, grounded in the brand kit and the recent posts.

  3. 03

    It checks its own work

    A guard rejects any invented number, price, or claim and regenerates before a person ever sees the draft.

  4. 04

    Approve, or send it back

    The draft lands on a review board. Drag it to approved and it schedules itself. Drag it to revision with a note and it rewrites itself from that note.

  5. 05

    It publishes

    Approved posts go out to Facebook, Instagram, and LinkedIn on their own, with the live links written back to the board.

· Results

Generates
Caption + images
Human does
Plan + approve
Status
Live in production

1. The studio generates both the caption and the images for each post with a brand-aware model, and a guard checks every line against the brief, rejecting invented stats, prices, or quotes and regenerating before a human sees it.

2. Revision is a closed loop. A reviewer drags a card to revision and leaves a note, and the post rewrites itself from that note, with no one opening an editor.

3. Approved posts publish to Facebook, Instagram, and LinkedIn through a claim-based worker that never double-posts and writes the live links back. Spending is capped at four levels with a live meter and a kill switch, and the whole thing is multi-tenant, deployed as white-label client instances.