The big idea: Human in, human out. AI amplifies the strategy, judgment and voice you bring to it, and the gap between what AI writes and what you rewrite is your brand voice.
See it in action
Image Prompt Lab
Build an image prompt one part at a time, the way you would brief a photographer. Watch the preview change, then regenerate to see what the tool guesses when you leave a part out.
Before you use any AI image
- Iterate on purpose. Change one part at a time so you know what made the difference.
- You are the judge. Check hands, text, product details and brand fit. Ask: could any brand have posted this?
- Rights and disclosure. Read the tool’s terms for commercial use. Never prompt a living artist’s name, a real person or another brand’s logo. In the U.S., images made entirely by AI generally can’t be copyrighted, so your human edits matter. Label realistic AI images where the platform or client asks.
Key concepts
- Human on the front end: AI doesn’t know your brand. Give it your positioning, audience, five personality words, goal and constraints.
- Human on the back end: AI output is a draft and you’re the editor. Ask: could any brand have written this?
- Match the tool to the task: Different AI tools have different strengths for voice, volume, research and trends.
- Run one prompt through two tools: Comparing outputs makes a usable result much more likely.
- Unclaimed territory: Use AI to map competitors’ style and tone, then look for what nobody owns yet.
Real examples
- Heinz “Draw Ketchup”: Ask AI to draw ketchup and it draws something like a Heinz bottle. AI proved a human brand truth; it didn’t create it.
Real brands in action
Heinz: Draw Ketchup
In 2023, Heinz asked AI image generators to draw “ketchup” with no brand name and no other hints. Nearly every image looked like a Heinz bottle. Heinz did not lead with the technology. It used AI to prove something people already felt: when the world thinks ketchup, it thinks Heinz.
The lesson: Start with a human insight, then use AI to prove it. The idea still comes from you.
Learn more: Heinz
Coca-Cola: AI extended the brand instead of inventing it
In 2023, Coca-Cola launched Create Real Magic, an open platform where creators remixed its iconic assets with DALL-E and GPT-4. The strategy stayed human. AI amplified what creators brought to it, and the brand’s decades of visual equity kept the work on brand.
The lesson: Treat AI as an extender of your ideas, not the source of them.
Learn more: Coca-Cola
Brand images are shown for educational purposes only. Each image and trademark belongs to the brand credited, and each credit links to that brand.
Frameworks from class
The editor’s mindset
AI output is a draft and you are the editor. Read every draft with three questions before you keep any of it. If you submit AI output unchanged, you have stopped doing marketing.
- Evaluate: does this sound like us and reflect our positioning?
- Challenge: is the argument sound, are the claims accurate, would our audience believe it?
- Make it yours: add a specific example, a sharper phrase, or a point of view only your brand could own.
Know each tool’s defaults
No AI tool is the best at everything. Each was trained differently, so the same prompt comes back different. Learn what each one does well and where it lets you down.
- Claude: brand voice, personas, positioning. Great depth, lower volume.
- ChatGPT: fast drafts, lists, structured formats. Generic by default, so push it with specific words and tone cues.
- Gemini: live web content, YouTube and Google tools. Less steady at holding a brand personality.
- Grok: real-time trends and cultural references. High energy, weak on strategy.
Insight first, AI as proof
Heinz did not ask AI for an idea. The team already knew that when people think ketchup, they think Heinz. AI was the experiment that made that truth visible. Use the same loop: human insight, then AI execution, then human judgment.
- Start by naming a human truth you already know about your brand.
- Use AI to prove it, not to discover it.
- Keep the concept simple. Clarity beats complexity.
Tools come and go. Judgment stays.
Next year there will be new platforms and new features. What you build now is the habit of adapting. AI can generate, but it cannot understand your audience’s deepest needs or build a brand story. It does not know your client’s unspoken fear or choose a word because of how it feels. That is where your value as a marketer lives.
- Chain tools in one workflow: brainstorm in a chat tool, lay it out in Canva, polish with Magic Studio, then experiment with an image tool.
- Treat AI like a fast creative intern. You are still the strategist making the calls.
Class exercise: Competitive landscape audit
Use AI to map your competitors, then add what only your team knows.
- Prompt an AI research tool: I’m building a brand called [NAME], a [CATEGORY] brand for [AUDIENCE]. Name the 5 most recognizable competitors and describe their visual style, tone of voice and the emotional territory they own. What territory is not being claimed?
- Review the output as a team. If it feels generic, push back and ask for niche competitors or more on emotional territory.
- Write down one insight the AI got right.
- Write down one gap that feels real and actionable for your brand.
- Write down one thing it missed, such as local context or insider knowledge. That friction is often your best insight.
Try it
Write one brief with your brand name, category, audience, five personality words and positioning, and ask for an Instagram bio, a tagline and a launch caption. Run it through two AI tools, score each piece 1 to 5 for brand fit, circle the best line, then rewrite the weakest piece yourself.
Tips and best practices
- Write the brief before the prompt. Know the audience, the job of the image and the brand’s personality words first. A prompt is only as good as the strategy behind it.
- Name every part of the shot. Subject, setting, style, lighting, composition and mood. Anything you leave out, the tool decides for you, and it decides differently every time.
- Change one thing per round. Save each version with its prompt. When something improves, you will know which change did it and can reuse it.
- Add text and logos yourself. Image tools still misspell words and warp logos. Ask for empty space, then set headlines and the logo in Canva, Figma or Adobe Express.
- Inspect before you publish. Zoom in on hands, faces, product details and backgrounds. Then ask whether any other brand could have posted the same image.
- Check rights and disclosure every time. Use tools whose terms allow commercial use, never prompt a living artist’s name or a real person, and follow the platform’s and client’s rules for labeling AI images.
Quick check
Think through your answer, then select a question to reveal one.
What does “human in, human out” mean?
People supply the strategy going in and the judgment and editing coming out. AI works in the middle.
What question should you ask before publishing AI-assisted copy?
Could any brand have written this?
Go deeper
- Writing and strategy: Claude, ChatGPT
- Visual ideation: Midjourney, DALL-E, Adobe Firefly
- Research: Perplexity, Gemini
View the AI as a Creative Partner slides on Gamma