Marketing agencies have been exploring Generative AI possibilities since ChatGPT came out (or even since Dall-e was introduced), and it seems like some of them are now looking for automated content creation tools. Over the past few months, I’ve received remarkably similar pitches from 2-3 different marketing companies and founders, all centered around the same core concept.

While each request had its unique twist, they all boiled down to the same fundamental idea: building an AI system that can automatically generate image-posts for social media, to replace what Graphic Designers do on tools like Canva, but with the possibility of editing the image-posts later, either with more prompting, or manually (like a Designer).

Let me share what these requests looked like and why they represent a problematic approach to AI product development.

The Project Requests

The conversations typically started something like this:

“We want to build an AI-powered social media image-posts generator. Think Canva, but automated. The AI should be able to create professional-looking social media image-posts with text overlays, brand colors, and layouts. The key requirement is that the generated images must be editable afterward, especially the text elements. Users should be able to either edit manually or use prompts to make changes.”

As the discussions progressed, the requirements usually expanded to include:

The Problems

The fundamental issue with these requests isn’t that the technology is impossible, but rather that the approach ignores critical technical limitations of current AI image generation systems, and that A.F.A.I.K an out-of-the-box technology that we could use, doesn’t actually exists (once again, as far as I know).

The Layer Problem

Current AI image generation models create single, flattened images. They don’t generate multiple editable layers like professional design tools require. When you generate an image with text, background, and graphics, everything is baked into one layer.

For editable social media posts, you need:

• Separate text layers that can be modified independently
• Background elements that can be swapped or adjusted
• Logo and brand elements on their own layers
• Graphic elements that can be repositioned or resized

The only viable solutions are either waiting for Canva to develop and integrate this capability into their existing platform, or building a complete Canva clone from scratch where the AI has granular control over each layer it generates separately.

The Logo Reproduction Problem

When clients request brand consistency with logos and brand elements, current diffusion models face a critical limitation: they don’t “copy and paste” existing logos into generated images. Instead, they attempt to recreate the logo during the generation process.

This approach consistently introduces:

• Visual artifacts and distortions
• Color variations from the original brand guidelines
• Shape modifications that compromise brand identity
• Inconsistent rendering across different generations

Diffusion models are trained to generate new content, not to preserve exact visual elements. This makes reliable brand reproduction nearly impossible with current technology.

Alternative Approaches

Instead of trying to build an AI-powered Graphic Designer replacement, here are focused approaches that work within current technical limitations:

1. Smart Asset Recommendation Engine

Build a tool that analyzes content requirements and recommends appropriate stock photos, icons, and graphics from existing libraries. The copy (text) can be easily generated by LLMs (as almost everyone already knows how to do that). The AI understands context and brand guidelines to surface relevant assets that can then be composed in traditional design tools.

Why this works: Speeds up the design process without trying to solve the layer generation problem, or automating the entire design process.

2. Template-Based Content Assembly

Build a system that combines AI-generated copy with pre-designed templates and stock assets. Users input their content requirements, and the AI selects appropriate templates, generates copy, and populates predefined areas with relevant stock imagery.

Why this works: Separates content generation from visual design, working within existing design frameworks while leveraging AI for what it does well: text generation and asset matching.

3. Brand-Consistent Background Generator

Focus specifically on generating branded background patterns, textures, and layouts that complement existing logos and text elements. The AI creates visually cohesive backgrounds while leaving text and branding elements to be added via traditional design tools.

Why this works: Solves a specific design need without attempting full layout generation, and backgrounds can remain as single layers without editing requirements.

4. Content Variation Engine

Create a tool that takes one approved design and generates multiple content variations by intelligently swapping text, adjusting color schemes within brand guidelines, and suggesting different messaging approaches for the same visual layout.

Why this works: Maximizes the value of existing design work while using AI for content ideation and brand-compliant variations.

5. Social Media Copy and Caption Optimizer

Build an AI system that analyzes successful social media posts in specific industries and generates optimized copy, hashtags, and posting schedules. Pair this with simple design recommendations for existing tools.

Why this works: Focuses on the content strategy layer where AI excels, leaving visual design to established tools while still providing significant marketing value.

A Realistic Path Forward

Rather than building a monolithic AI design tool, successful projects in this space should acknowledge the current technical limitations and work within them:

• Accept the layer limitation. Current AI cannot generate true multi-layer designs. Any solution must either work with flattened images or integrate with existing design platforms that handle layer management.

• Focus on workflow integration. Instead of replacing design tools, create AI assistants that work alongside existing platforms, providing suggestions, automating repetitive tasks, or generating initial concepts that designers can refine or use as inspiration.

• Plan for hybrid solutions. The most viable approach may be combining AI generation for copy generation, and for image generation of backgrounds and layouts with traditional asset placement for logos, images, and brand elements.

The goal should be augmenting existing design workflows with AI capabilities that respect both technical limitations and professional design practices. Do not try to replace Graphic Designers, but rather augment their work, and make their job easier, faster, better, more efficient, or even all of them at the same time.

If you’re considering an AI project in the marketing or design space (or any other field), I’m available for consulting to help you identify realistic, valuable applications of AI technology that work within current technical constraints.