AI isn't just about one tool anymore. It's about building smarter systems — and I discovered that firsthand while designing conference notepads.
I've been using OpenAI's ChatGPT quite a bit lately, both for image generation and for answering general questions. The image generation has improved dramatically. One thing I've noticed recently is that ChatGPT can now make small edits to images it already created — something it really struggled with just a few months ago.
I recently used it to help students in my adult school podcasting class. I wrote prompts describing their podcast videos, including the topics, style, and tone, and ChatGPT generated impressive YouTube cover art for them. The results were much closer to professional podcast thumbnails than what AI tools were producing even a year ago.
But I also discovered something important about AI workflows: one tool usually isn't enough anymore.
Using AI to Design Conference Giveaways
My husband Richard is speaking on a panel at the BioNJ conference, where we also have a vendor table for Gearhart Law. I asked ChatGPT for ideas for conference giveaways, and we decided on branded notepads.
Then I asked ChatGPT to design the notepad. The design included space for a QR code that would link people to our website. I created the QR code separately using an online QR code generator and then asked ChatGPT to place the QR code into the design.
Everything looked perfect. Until I tested the QR code. It didn't work.
At first, I thought maybe I had exported the wrong file format. I tested JPEG versus PNG versions. I checked the QR code directly on the QR code site, and it scanned correctly there. But once it passed through the AI image generation process, the QR code stopped functioning.
Why AI Breaks QR Codes
AI image generators often "reinterpret" QR codes as visual elements instead of preserving the exact machine-readable pixel structure required for the code to function. Even tiny visual alterations can break the embedded URL connection. This is a known limitation — AI sees patterns, not data.
The fix was simple once I understood the problem. Instead of asking AI to merge the QR code into the image, I uploaded the AI-generated design into Canva and added the QR code manually there. That preserved the QR code functionality. I tested it again on my phone before uploading the final file to an online notepad printing company, and this time everything scanned correctly.
The Real Lesson: AI Workflows Are Becoming Multi-Tool Workflows
What struck me afterward was that I had used multiple tools to complete one relatively simple project:
Generated notepad design concepts and conference giveaway ideas
Created a working QR code linking to gearhartlaw.com
Added QR code manually to preserve machine-readable pixel structure
Uploaded final file to notepad printing platform
That's becoming normal now. Instead of relying on one AI platform to do everything, many of us are starting to assemble workflows where different tools handle different parts of the process. And honestly, that's starting to look a lot like building lightweight AI agents.
Why This Matters for Business Owners and Content Creators
A lot of people still think of AI as "ask one question, get one answer." But increasingly, the real power comes from connecting tools together into repeatable systems. For example:
The businesses that learn how to connect tools together — instead of expecting one platform to do everything perfectly — are going to move much faster. I'm finding myself naturally building these mini-workflows almost every day now, whether it's podcast production, marketing, conference prep, or content creation.
So yes, AI image generation is getting better very quickly. But the bigger shift may be this: we're moving from using AI tools individually to building AI-powered systems. And that's why it's probably time to start building AI agents.
TL;DR — Key Takeaways
- ChatGPT's image generation is now close to professional quality for thumbnails and marketing graphics
- AI image generators break QR codes — always add QR codes in Canva or Photoshop after AI generation
- Real projects now require multiple AI tools working in sequence, not one platform doing everything
- Multi-tool workflows are the foundation of AI agents — systems that complete tasks automatically
- Businesses that build repeatable AI workflows will outpace those using AI one prompt at a time
FAQs About AI Image Generation, QR Codes, and AI Workflows
Why do QR codes sometimes stop working after using AI image generators?
AI image generators often modify the exact pixel structure of QR codes while trying to 'improve' or blend the image visually. Even tiny changes can make the code unreadable to scanners. The fix is to add the QR code after AI generation, using a design tool like Canva that preserves the exact pixel structure.
What's the best way to add a QR code to an AI-generated image?
Generate the design first with AI, then add the QR code afterward in a design tool like Canva or Adobe Photoshop. Avoid sending the QR code itself through the AI image generation process, as AI will reinterpret it as a visual element and break its machine-readable structure.
Can ChatGPT generate YouTube thumbnails and podcast cover art?
Yes. ChatGPT can now generate highly customized images and YouTube-style thumbnails from detailed prompts. Results improve significantly when prompts include the podcast topic, mood, colors, target audience, and visual style.
What are AI agents?
AI agents are systems that combine multiple AI tools and automated steps to complete tasks with minimal human involvement. For example, an AI agent could create content, generate graphics, publish social posts, and send follow-up emails automatically.
Why are businesses using multiple AI tools instead of one?
Different AI tools specialize in different tasks. One may be better at writing, another at graphics, another at automation, and another at video editing. Combining them into a workflow often produces significantly better results than relying on a single platform.
Are AI-generated graphics good enough for professional marketing?
Increasingly, yes — especially for social media graphics, YouTube thumbnails, podcast covers, presentations, and promotional materials. But final human review is still important, especially for branding accuracy, text readability, and functional elements like QR codes.
What's the biggest mistake people make with AI-generated marketing materials?
Many people assume AI-generated output is ready to use immediately. In reality, AI content often needs testing, editing, formatting, or refinement before it's production-ready. Always test functional elements like QR codes, verify brand accuracy, and review text for errors before printing or publishing.
How can small businesses start building AI workflows?
Start simple: identify repetitive tasks, choose one AI tool for each task, connect the steps into a repeatable process, and gradually automate pieces over time. Most businesses don't need complex AI systems to benefit — even simple multi-tool workflows can save substantial time.