AI Social-Content Engine

How can AI generate authentic, platform-specific social media content while preserving a brand's identity and creative direction?

Creating content for social media requires much more than producing attractive visuals. Every platform has its own style, pacing and audience expectations, while every brand has its own visual identity and tone of voice. I wanted to explore whether AI could go beyond simply generating prompts and instead think more like a creative team—developing strategy, understanding products and producing content tailored for different platforms.

Rather than relying on a single prompt, I designed a multi-agent system where each agent performs a specific creative responsibility. One agent develops a social strategy from the brand brief, another analyses each product, while specialised agents generate image and video prompts designed specifically for platforms such as TikTok, Instagram Reels and YouTube Shorts. The system also uses product reference images rather than text descriptions to preserve product accuracy, allowing prompts to focus entirely on composition, lighting, storytelling and atmosphere.

Agent 1 – Brand & Social Strategy Develops the brand's visual identity, content strategy, platform differences, creative direction and content pillars. Agent 2 – Product Analysis Understands each uploaded product and its role within social media content. Agent 3A – Social Image Prompt Writer Creates platform-ready image prompts using product reference images. Agent 3B – Video Prompt Writer Generates image-to-video prompts and storyboard outputs with explicit timing instructions.

1. 01

   User submits a brand brief.

2. 02

   Uploads product images.

3. 03

   Optionally uploads moodboards and previous social content.

4. 04

   Agent 1 develops a creative strategy.

5. 05

   Agent 2 analyses each product.

6. 06

   Image and video agents generate production-ready prompts.

7. 07

   Prompts can be downloaded individually or as a ZIP archive.

• —

  One of the biggest technical challenges involved deployment. Railway's multiple Gunicorn workers caused jobs to disappear because requests could reach different processes. This was solved by reducing the application to a single worker.

• —

  Creative quality required significant iteration. Early outputs looked overly polished and artificial, leading me to redesign the prompt architecture around authenticity, natural moments and lifestyle storytelling.

The final system successfully generates platform-aware social media prompts for both images and short-form video while maintaining brand consistency. Testing with brands such as Poppi and Aritzia demonstrated that the multi-agent workflow could produce outputs that felt significantly more authentic after several prompt architecture refinements.

This project shifted my perspective from building AI prompt generators to designing AI creative systems. The biggest lesson wasn't how to write better prompts—it was learning how to divide creative work into specialised responsibilities. Giving each AI agent a clear role produced significantly stronger outputs than asking one model to solve every problem at once. It also reinforced that successful AI systems rely just as much on creative direction and workflow design as they do on the underlying model itself.

• →

  Persistent brand memory

• →

  Campaign history and idea tracking

• →

  Trend research integration

• →

  Performance analysis

• →

  Social media dashboard

• →

  Content scheduling

• →

  Creative feedback agent

• →

  Caption and voiceover generation