Give it a topic then ask for a campaign and AI can hand you something that looks ready to use before you have decided what you want to say. Accept that result too quickly, and you can skip the work that would have helped you discover a better idea.
That is the risk I see in much of the AI work that crosses my path. An image filter produces an impressive picture, or a prompt produces an app that closely resembles something already available. The result may be useful and the execution may take skill, but the finished product alone tells me little about the creative thinking behind it.
I use AI extensively, so I have to examine my own habits, too. A short prompt might follow weeks of thought, and a familiar idea can become something valuable through a fresh interpretation. What happens when getting a result becomes so easy that I stop exploring what else the work could become?
Creativity needs room to develop
Growing up, I never thought I was creative. That changed when I began to understand how to develop creativity in myself. I had to discover that I could contribute ideas, rather than assume that ability belonged to someone else.
Too much trust in AI gives me an easy way to skip that discovery. If I already believe I’m not creative, a polished response gives me little reason to test that belief. I can finish the project without ever finding out what I might have contributed.
In The Fallacy of AI: Why It’s Brilliant But Still Needs You, I explored the limits of what AI could produce from its available context. I asked for a mug with coffee leaking through holes. I also asked it to envision a device combining a phone, music player, and web browser using only context reasonably available in 2005. In both cases, I was testing whether AI could work beyond familiar representations to produce the unfamiliar result I was asking it to imagine.
Those experiments raised a question that matters beyond the images themselves: how much should I let AI's output define what I believe is possible? If it cannot produce an idea, I can keep exploring that idea myself. But if I expect AI to originate the idea in the first place, I may never consider a direction it does not offer.
That is where the limits of the tool can become limits I place on myself. Creativity needs room to develop through questions, experiments, and connections I have not made before. AI can participate in that exploration, but accepting its suggestions as the full range of possibilities could keep me from discovering new ideas or recognizing my own creative ability.
(This image of my mug shows how far AI image generation has come in the few shorts months since my Fallacy of AI post. While it's come a long way at creating something outside of what's been created before, it still has a bit to travel. Or, maybe it's just seen my other post, knows how much I comment on it's inability and wants to show me up.)
AI can improve an output and still narrow the possibilities
The research complicates any blanket claim that AI makes people less creative. In an experiment by Anil Doshi and Oliver Hauser, access to AI-generated ideas helped participants produce stories that received better evaluations, including for creativity. However, the AI-assisted stories were also more similar to one another than the stories written without that assistance. The experiment evaluated the stories; it did not establish whether the writers developed their own creative ability over time.
That finding comes from a particular writing experiment, so I would not stretch it into a verdict on every designer, marketer, or developer. It does show why a better individual result and a wider range of ideas are different questions. A team can improve the finish of its work while still needing to ask whether it is exploring enough different directions.
A separate study of 319 knowledge workers found that greater confidence in generative AI was associated with less reported critical thinking. That was a survey of people's experiences, not proof that AI permanently reduces thinking ability. I take it as another reason to examine how readily I accept an answer that sounds convincing.
When AI gives you a convincing answer, do you keep exploring or start polishing? Once something looks finished, it is easy to focus on making it sound better. Before investing that effort, ask whether the idea itself deserves to go further.
Collaborate with AI before asking it to finish
Writing this blog post is a pretty good example of what I mean. Before asking for my first draft, I talked through the idea with AI. I wanted to understand where my argument held up, what I might be overlooking, and whether I could explain it clearly enough to make it useful.
Here’s how that collaboration carried the idea from an initial hypothesis into the draft I’m editing now:
- State the hypothesis. I began with the idea that AI can stifle creativity by making it easy to skip the thinking that helps us develop it. That gave the conversation a position to examine.
- Test the idea through discussion. I described how I had once thought of myself as uncreative. AI challenged the idea that people fall into fixed creative and noncreative groups. That helped sharpen the question: could getting a finished result too easily keep someone from discovering and developing their own creativity? The discussion gave me a clearer argument to investigate, rather than proof that my starting position was correct.
- Draft with the discussion and research in mind. Only then did I ask for a first draft. The research challenged me to distinguish a better creative result from developing the person’s creativity. I could keep the bold opening while making that distinction clear.
- Keep collaborating while editing. When the draft linked to my earlier article about AI's limitations, the connection felt forced and misrepresented why I had used the coffee mug example. I explained what the example was meant to demonstrate, challenged the transition, and asked for the whole section to be reworked. I was still developing the argument while deciding how to express it.
That follows the pattern I found in my review of more than three years of AI work: set direction, inspect the response, challenge what does not fit, and revise selectively. As I edit this article, I am applying that process to the claim the article makes. AI can help me explore, research, draft, and revise, while I remain responsible for deciding what I believe and what is ready to publish.
Put that collaboration to work in your marketing
You can use the same approach when developing anything. Consider an example of creating a marketing campaign for your company.
Imagine your business wants more website inquiries. Conversations with potential customers might suggest that people are unsure what happens after they reach out. Your starting hypothesis might be that explaining the first conversation would make contacting you feel easier.
Before asking AI to build a campaign around that idea, use questions like these to test your reasoning:
- What assumptions am I making about why people hesitate to contact us?
- What other explanations could fit the customer feedback I’ve shared?
- What is the strongest argument against my idea that explaining the first conversation would help?
- What could I ask potential customers to find out whether that explanation is right?
- What evidence would suggest I should change direction instead of building this campaign?
Now bring a creative direction of your own into the conversation. You might propose showing the first ten minutes of the initial conversation, letting a potential customer see the questions you ask and what happens next. Ask AI to challenge that approach and suggest other ways to make the experience understandable, then develop, combine, or reject those ideas based on what you know about your customers.
Once you have a direction worth developing, AI can help with the copy, visual concepts, or page layout. Keep checking whether the draft expresses the idea you chose, and sketch or explain a missing element yourself if the tool struggles with it. You do not have to abandon a useful idea simply because AI's first attempt fails to capture it.
Then show a rough version to someone who resembles your intended customer. Ask what that person expects to happen after making an inquiry and what still feels unclear. Use the response to revisit your hypothesis, develop the idea further, or change direction before investing in finished production.
You own the campaign. You own the creative direction. You own the results. AI helps with research, offers counterpoints, and handles much of the production work.
Keep a part of the work that makes you think
I don't measure my creative growth (or worth) by how much work I refuse to let AI do. I look at the decisions I can explain afterward and whether I understand the problem more clearly. That connects to learning with AI without giving up the thinking: an impressive answer is only part of what you can gain from the experience.
Choose one project you are working on now and name the observation behind it, the alternatives you considered, and the reason you chose this direction. If those answers are difficult, return to the idea before requesting another polished version. You may find that AI is more useful when you ask it to help you investigate the uncertainty.
AI can stifle creativity when it becomes a convenient place to stop thinking. It can also help you explore further when you keep asking questions and making informed choices. On your next project, what decision will you keep for yourself so you come away with more than a finished artifact?
