AI, Creativity, and the Risk of "Meh"
As AI makes polished work easy to produce, designer and educator Debbie Millman says the real risk is everything becoming "pretty good, or meh," and brands with no point of view.
Topics
“The greater danger is that everything becomes pretty good, or meh.”
That is Debbie Millman‘s worry about what AI is doing to creative work. Bad work, she says, at least shows effort. Her fear is a flood of competent work in which nothing stands out.
It is a pointed warning for marketers, who can now produce polished language, imagery, and strategy quickly, often with the same tools as their competitors.
Millman is a designer, author, educator, and strategist who has hosted Design Matters for more than two decades. She uses AI for research but not to write her interview questions. She is an executive fellow at Harvard Business School and sits on Canva’s design advisory board.
Her argument is about sequence, not rejection: use AI after you have built the judgment to direct it. For brands, the same holds for conviction. Tools can express an identity, she says, but only if one exists.
Excerpts from the interview;
Creativity often starts with permission to make something badly. What happens to that phase for a student who can get a polished result from AI on the first try?
I think we risk confusing polish with accomplishment. One of the most important aspects of becoming a creative person is making things that aren’t very good right out of the gate. You have to develop the capacity to recognise the distance between what you intended and what you actually made. That distance is where taste, judgment and craft begin to develop.
AI can collapse that distance almost instantly. A student can get something that looks finished before they have developed the ability to understand why it works, whether it works, or whether it says anything they actually believe.
I say this as someone who uses AI for research and, more recently, for search. But I’ve spent decades developing a point of view and a practice, and I am still doing it. In that sense, I feel comfortable using AI as a tool for leverage because I spent years doing the work without it. My concern is what happens when we give people the leverage before they’ve built the muscle.
AI can help a non-native English speaker compete on a more equal footing without replacing their voice, but the line between an equaliser and a shortcut is thin. What signals tell you that someone has crossed it?
For me, the question is whether AI is helping someone express an idea they already have or supplying the idea itself.
If a student knows what they want to say and AI helps them with grammar, syntax, vocabulary or structure, that can be enormously empowering. Language proficiency and intellectual ability are not the same thing, and technology can help remove a barrier that may have little to do with the quality of someone’s thinking.
The line gets crossed when the student can no longer account for the thinking. If I ask why they made a particular argument, chose a certain word, reached a conclusion or structured something in a particular way, they should be able to tell me. I become concerned when the language is considerably more sophisticated than the thinking underneath it: the writing sounds polished on the surface, but it’s really a string of words that may sound good together.
I am also concerned when someone can’t explain or defend what they have supposedly written. The goal should be greater access to one’s own voice, not outsourcing its development.
You use AI for research on Design Matters but not to write your questions. Can you share an example of a question that came purely from you, and why it mattered?
One of my favourite examples comes from my interview with Michael Arden. I had learned that when he was four, his grandmother took him to see Sesame Street Live. I wanted to begin there, not with his Broadway career or his Tony Awards, but with this very small, almost incidental fact from childhood.
What interested me was the anecdote’s specificity. That is the kind of detail I look for because it belongs uniquely to that person. And when we began talking about it, a remarkably rich story emerged.
AI is extremely useful to me for locating facts, tracing chronology and finding obscure interviews. But deciding that this tiny fact is where I want to begin is the creative act for me, and I would be loath to let that go. Questions crafted by AI aren’t fueled by curiosity. When I’ve experimented, the results were questions without “soul,” literally and figuratively.
A good question comes from intuition, curiosity and everything I have learned from 21 years of sitting across from people and listening very carefully to how a life reveals itself.
After two decades of interviewing creative people, what common thread have you noticed in how they work? Does AI make it more or less important?
One word: Curiosity. The people I find most compelling remain extraordinarily curious. They keep looking, reading, listening, questioning and paying attention long after they have achieved whatever the world might define as success.
Curiosity is different from having access to information, and we now have access to information that is almost unimaginable. AI can answer questions at extraordinary speed. But it cannot decide which questions you should ask or which ones really matter (no pun intended).
The creative people I admire most are not merely good at producing things. They are good at noticing things. I don’t see AI making that capacity obsolete. I think it may make it one of the most important creative abilities we have and can still provide.
You sit on Canva’s design advisory board and think carefully about where AI helps or hurts creative work. Is there a recent example of feedback you gave on a feature that sparked a good back-and-forth?
I’m careful about discussing specific product details publicly that may be confidential, but one recurring conversation I find most interesting is about the difference between making creativity easier and making creative outcomes more generic.
I am enormously enthusiastic about tools that eliminate unnecessary friction. But there is a danger when ease begins to remove decision-making. Every meaningful creative act involves choices: Why this image? Why this word? Why this typeface? Why this composition? Why this idea rather than another one?
So the question I tend to bring into these conversations is: Where is the human decision still happening? The most interesting exchanges occur when we stop asking only what AI can generate and start asking what role the person is being invited to play.
When AI can produce and polish brand communication, what makes a brand memorable, and what makes it feel interchangeable?
A memorable brand has a unique, single-minded, purposeful, original, soulful point of view. AI can sometimes produce competent language, imagery and strategy at what seems like the speed of light. What it can’t manufacture on its own is conviction.
It cannot decide what an organisation is willing to stand for, what it will refuse to do, what risks it will take, what experiments it will undertake or what it believes strongly enough to remain consistent about when circumstances become difficult. That is where brands become distinctive.
The danger with AI isn’t necessarily that everything will become bad or bland. I think the greater danger is that everything becomes pretty good, or meh. In my mind, I’d rather something be bad than mediocre. At least something bad shows some effort and a decent attempt.
The brands we remember tend to make choices and take risks. They have a recognisable purpose and values, clear and single-minded language, and telegraphic equities. They leave things out and sometimes alienate people.
And because branding is still, foremost, an act of sacrifice, they understand that identity is created as much through exclusion as through inclusion. AI can help a brand express itself, but first, there has to be an actual self to express.
ALSO READ: Fairing CEO Matthew Bahr on the Limits of AI Attribution
