AI Can Make It Look Real. That Doesn't Make It a Product.
AI has done something pretty extraordinary for design: it has made the first step look remarkably like the last.
Give it an image of your muse. Tell it about your target customer, your aesthetic, your colors, the category you're designing, and some of the details you want. Keep prompting and refining, and before long you can have what looks like a beautifully designed collection, photographed on beautiful people in beautiful locations.
But did you actually design it?
I use AI all the time, so this is by no means an argument against it. It is an amazing design tool, research tool, and thinking partner. But particularly for founders developing physical products, we need to be very clear about what it can do and what it only looks like it has done.
Because creating an image of a product and creating a product are two very different things.
Can AI design my collection for me?
It can certainly help you visualize one.
Feed AI enough information and it can collect, synthesize and reinterpret it into a compelling surface-level point of view. But especially in fit-critical categories like lingerie, swim and activewear, design isn't surface deep.
A garment is three-dimensional. It has weight, stretch, recovery, support and construction. It has to move with the customer, work on different bodies and across a range of sizes, be worn hour upon hour, washed repeatedly, and still perform the way you intended.
You can tell AI that you want a stretch fabric. But what stretch fabric? With what weight and recovery? How much compression? Does it work with the construction you're imagining? Can your factory work with it? Can you afford it at your target price? How will it behave as the garment is graded into different sizes?
The image can skip every one of those decisions. The product cannot.
And there is another limitation we don't talk about enough. An LLM has access to an enormous amount of information created by designers, developers, production people, founders, and countless others. What it doesn't necessarily know is who among them is qualified to be repeated, or how the end product actually turned out.
It hasn't stood in the fitting room when something that should have worked didn't. It hasn't received the sample that looked nothing like the sketch. It hasn't watched a fabric lose recovery after washing or realized that one small pattern adjustment created a problem somewhere else.
Experience is not simply knowing what should work. It is having seen what actually did.
But if it looks good, why does it matter who designed it?
It doesn't.
I don't particularly care whether you call yourself a designer, whether you hire one, or what title you give the person helping you. What matters is the product's integrity.
Your customer isn't buying your rendering. She is buying the garment.
She experiences where it hits her body. Whether the waistband rolls. Whether a strap stays in place. Whether the fabric feels good after six hours. Whether the garment she loved the first time she put it on still performs after five washes. And she experiences all of that on a body that may be very different from the one shown in your image.
This is why I don't believe you own the design simply because you've created a pretty picture. You begin to own it when you do the work required to turn that idea into something real.
That work is also where innovation happens.
Innovation doesn't necessarily mean a revolutionary, patented new, technology. Sometimes it does. But it can also mean solving a problem differently: an unexpected closure, a different construction technique, a new application for a material or a seemingly small change that makes the product work substantially better for the customer.
Those ideas come from problem-solving, curiosity, and experience. You try something. You see what happens. You ask why. Someone else on the team sees something you didn't. You change it and try again.
Innovation rarely happens by staying on the surface.
When the first step looks like the last step
I've been sketching garments for most of my career. Some of those initial illustrations can be quite beautiful.
Nobody has ever confused one with a finished garment.
Everyone understands that the sketch is the beginning of an idea we're going to develop. And even that sketch isn't static. My original illustration may change five to ten times during the flat-sketch stage alone as I begin thinking through how the garment actually works. Often, as the design develops and bigger problems emerge, that pretty illustration becomes redundant.
What happens to this seam? Where does that strap go? Does this line connect logically to the back? Can that construction actually be made? Do I still like the proportion from another angle?
An AI-generated image hasn't necessarily had the benefit of any of that exercise.
And that, to me, is one of the biggest differences AI has introduced into the design process.
The first step can now look almost indistinguishable from the last.
You can take a concept that hasn't been patterned, sourced, engineered, sampled, fit, graded, wear-tested, washed, costed, or produced and put it onto a seemingly real person. The garment can have shadows, folds, and texture. It can sit against the body convincingly enough that you start to believe it exists.
Visually, you've jumped from Step 1 to Step 10.
The product hasn't.
An image is static. A product has to live on the body.
I learned this lesson early in my career, and frankly, I still encounter it.
I can design a front I love and a back I love and then realize when I start working through the garment that the two don't actually work together.
A strap has to go somewhere. An armhole affects something else. Lines that look beautiful from the front need to resolve around the body and make sense from the back.
It's one of the reasons I still sketch on paper, move into flat sketches, and work on a form. Sometimes I'll create a very rough, Frankenstein version of an idea simply because I need to see how those lines travel around an actual three-dimensional body.
And I've been doing this for decades.
AI can give you a front, back, and side view. What it can't give either of us is the physical experience of discovering that what looked right in those images doesn't work when those three views have to become one garment.
The image is a signifier toward the collection. It is not the design.
Eventually, someone has to make it
This is where an inexpensive shortcut can become surprisingly expensive.
Let's say you create an AI image of a swimsuit you love and send it to a factory to make a sample. But you haven't completely thought through the fabric, construction, measurements, pattern, components, or technical details. You know what you want it to look like, but you haven't defined what it needs to be.
The factory has to fill in the blanks. Maybe that first sample costs you $300. It arrives, and your immediate reaction is, "That's not what I wanted."
But what exactly did you tell them you wanted?
If the information wasn't there, the factory made decisions based on its interpretation of an image. Now you make changes, pay for another sample, and another shipment. Perhaps the fabric wasn't right. Perhaps the factory wasn't right. Maybe it told you it could make swimwear, but its real expertise is ready-to-wear.
The supposedly inexpensive route starts getting expensive very quickly.
The same thing can happen with fit comments, grading, patterns, and construction. AI may even be able to look at a photograph of your first prototype and give you perfectly plausible-sounding fit comments.
But "plausible" is a fairly dangerous standard when those comments are about to drive your next physical sample.
Shortcuts don't necessarily eliminate the work. Often, they simply move the work downstream, where mistakes cost more.
And credibility matters
Founders should consider another piece. Your customer isn't the only person evaluating your product knowledge. Your factory is. So is a potential buyer, partner or investor.
Experienced vendors know when they're looking at an AI-generated concept. Using AI won't make them question you. Acting as though an AI-generated image represents a fully designed garment might.
Because someone will eventually ask the next question.
What fabric are you using? What is the construction? How does this work? What are the measurements? How should this be graded? What happens here when the body moves?
The more finished the image looks, the easier it is to forget you still need to know the answers.
So should you stop using AI?
Absolutely not. I certainly haven't.
There are so many places where I think founders should be using AI, and I'm going to get into those next. Research, competitive analysis, customer reviews, innovation, technology, trade shows, preparing better questions for experts, organizing information, and understanding terminology are only the beginning.
The issue isn't whether you should use AI. It's understanding what job you're asking it to do.
Use it to help you think, research, explore, work faster, and visualize possibilities. Just don't confuse its ability to make an idea look finished with the expertise required to actually finish it.
Your idea matters. Your taste matters. Your understanding of your customer matters. The energy and passion that made you want to build the product in the first place absolutely matter.
AI doesn't take any of that away from you.
But an image of your idea is still an image of your idea.
The image may be the beginning of the conversation. The product is what happens next.
Next week, we'll talk about where I think AI absolutely belongs in the process, and how founders can use it to become smarter, faster, and better at building real products.
If you're building an apparel brand and want to make sure your development strategy is working for your brand, not against it, that's exactly the kind of clarity we work through together. Book a call here to talk through where you are and what your product actually needs.
Comments