Thoughts on Design …

The first time it happened to me, I'll admit my stomach dropped. Not because the image was good. It wasn't. It was because the person sending it thought it was, and because it had cost them nothing. For a lot of my career, the hard part of design was making the thing. Now anybody can make a thing. That changes the conversation, and it changes it fast.

I've spent my working life as a designer and creative director, most of it in print: direct mail, appeals, annual reports, brand work, the kind of pieces that land on a kitchen table and have about three seconds to earn a second look. I've watched desktop publishing replace paste-up, watched the web eat a good chunk of print, watched stock photography go from a catalog on a shelf to a search bar. Every one of those shifts was announced as the end of the profession. None of them were. But I'd be lying if I said this one feels the same.

So this is my attempt to think it through out loud. Not as a futurist, and not as someone selling a course on prompt engineering. Just as a working designer who uses these tools, worries about them, and still believes the job is worth doing. Here's what I see changing, what I think is being lost, what I'm convinced isn't going anywhere, and what I'd tell a young designer, or the person hiring one, today.

Let's start with what's real, because a lot of the noise around AI in design is either panic or a sales pitch, and neither one helps anybody make a decision.

The tools are already in the room. They're not coming; they're here, and most of them showed up inside software we were already paying for. Photoshop's Generative Fill extends a background, removes a power line, or swaps a sky in the time it used to take to open the Clone Stamp tool. In June of this year, Adobe pushed its Firefly AI assistant into Illustrator and InDesign, the two programs I live in, with features for building brand kits, reorganizing layers, and catching missing fonts. Canva has Magic Studio. Figma has its own AI features. And sitting outside all of them are the image generators, Midjourney and the rest, that can conjure a photo-real scene from a sentence.

Businesses are using them, too. A Clutch survey published this spring found that 88 percent of companies use AI design tools to some degree. Canva's tools led the list at 43 percent, with Adobe Firefly close behind at 41. The jobs they trust AI with are telling: image editing and background removal, social graphics, ad variations, early concepts. In other words, the stuff that's fast, repeatable, and low-stakes.

Here's my take on using them day to day. Some of it is a gift. Extending a photo to fit a different aspect ratio used to be a half hour of careful cloning and a prayer. Now it's a click and a little cleanup. Pulling a subject off a busy background is nearly instant. Generating twenty rough directions for a thumbnail sketch session gets the bad ideas out of the way quicker. None of that makes me less of a designer. It makes me a designer with more time for the part that matters.

Some of it is a trap. Generated imagery has a look, a kind of airbrushed, too-perfect, slightly uncanny sheen, and people are learning to spot it. In nonprofit work especially, where the whole appeal rests on trust, a fake photo of a child who doesn't exist isn't a shortcut. It's a liability. I'll come back to that.

And some of it is just a better version of something we already had. Content-aware fill wasn't new. Auto-trace wasn't new. Stock libraries weren't new. AI is faster, broader, and stranger than any of those, but it sits on the same line of tools that were supposed to make designers obsolete and instead made them busier.

That said, I don't want to be the guy who waves all this away. Something is being taken, and it's worth naming plainly.

The first thing AI eats is production. Resizing a campaign into fourteen banner sizes. Cropping a hundred product shots. Writing six headline variants for an A/B test. Clipping paths, background knockouts, color matching across a set of photos. That work used to fill a junior designer's week, and a lot of it can now be done with a prompt or a batch action. If your value to a client was "I'll do the tedious stuff so you don't have to," the tedious stuff just got very cheap.

The numbers are mixed, which is about what you'd expect in the middle of a shift. The Clutch survey found that 18 percent of businesses say AI has reduced their need for designers. That's real, and it's not small. But 90 percent still use designers in some capacity, and nearly half said they'd increased design budgets over the past year. Meanwhile, the Bureau of Labor Statistics projects graphic design employment to dip about 2 percent over the next decade, from roughly 253,000 jobs to just under 249,000, with around 16,000 openings a year as people retire or move on. That's not a collapse. It's a slow squeeze.

The part that keeps me up at night isn't the total number of jobs. It's where the squeeze lands. Carly Ayres wrote a piece for It's Nice That last fall called "Welcome to the entry-level void," and the title says it. The work that used to be handed to juniors, the retouching, the resizing, the banner production, is exactly the work AI does best. Students are sending out portfolios and hearing nothing back.

Here's why that matters. Nobody walks out of school with taste. You get it the way I got it, and the way every designer I respect got it: by doing a thousand small, boring jobs under someone who would hand your work back and say, "Kern that again. Now look at it from across the room." The grunt work wasn't just labor. It was the apprenticeship. It was where you learned why a headline breaks where it does, why a reply device fails when the return address is in the wrong spot, why the photo of the smiling kid outperforms the photo of the sad one, or doesn't.

If we let AI take all of that away from young designers, we're not just saving money. We're eating our seed corn. In ten years we'll have plenty of software that can generate a layout and a shortage of people who can tell whether it's any good.

Now for the part I'm sure of.

AI can make a picture. It cannot tell you which picture to make. That sounds like a slogan, so let me be specific about what I mean.

Every real design job starts with a brief, and almost every brief is wrong. Not dishonest, just incomplete. The client says they need a brochure. What they need is for a skeptical donor to trust them enough to write a check. The client says they want it to look modern. What they mean is they're embarrassed by the last piece and can't say why. The client says "make the logo bigger," and what they're really telling you is that the board chair hasn't been heard in three meetings. A big part of this job is hearing the brief behind the brief, and no model sitting on a server is in that room.

Then there's taste, which is a word designers throw around a lot and rarely define. Here's my version. Taste is knowing what to leave out. It's the decision to set a headline in one weight instead of three, to let a photo breathe instead of filling the corner with a starburst, to cut the third bullet because two is stronger. AI is very good at producing more. It's not good at producing less, because less requires a point of view about what matters, and that comes from somebody who has watched real people respond to real work.

The businesses doing the hiring seem to know this, even the ones buying the tools. In that same Clutch survey, creativity was the trait they valued most in a designer, at 39 percent, followed by strategic thinking at 19. Speed and affordability each came in at 7. In other words, the very things AI is supposed to deliver are near the bottom of the list. What people want from a designer is the thing AI can only imitate.

And there's accountability. When a direct mail piece flops, somebody has to stand in front of the development director and explain why, and then fix it. When a brand mark accidentally looks like a competitor's, somebody has to have caught it. When a photo misrepresents the people it shows, somebody has to answer for that. Software doesn't carry that weight. A designer does. The more the machine generates, the more valuable the person who can sign off on it becomes.

I think of it the way I think about a good shotgun. A better gun doesn't make you a better shot. It helps, sure. But the lead, the swing, the follow-through, knowing when to take the bird and when to let it go, all that lives in the hunter. The tool extends what you already know. It doesn't supply it.

There are two questions I get asked more than any others now, and neither one is about craft.

The first is ownership. If we generate an image for a campaign, does the client own it? The short answer, at least in the U.S., is: not the way they think. In January 2025 the U.S. Copyright Office released Part Two of its report on AI and copyright, and it was fairly clear. Typing a prompt doesn't make you the author. Picking your favorite out of a batch of generated options doesn't either. A purely AI-generated image isn't copyrightable. What can be protected is the human part: the work a person creates and then modifies with AI, or a creative selection and arrangement of material.

I'm not a lawyer, and anyone building a brand on this should talk to one. But the practical takeaway for designers is worth stating. If a logo, a mascot, or a signature visual is generated wholesale, your client may not be able to stop a competitor from using something nearly identical. The safest place for AI is in the supporting cast, not the lead role. Draw the mark. Shoot or commission the hero image. Use AI to help with the background.

The second question is harder, and it's about trust. Should we be using generated images at all in work that's supposed to be true? For a lot of commercial work, I don't lose sleep over it. A stylized illustration of a coffee cup is a stylized illustration of a coffee cup. But I've spent a good part of my career on nonprofit and ministry work, where the whole piece rests on a promise: these are real people, this is a real need, your gift does a real thing. Put a generated child in a fundraising appeal and you've broken that promise, even if no one ever finds out. And increasingly, people do find out.

My rule is simple. If the image is standing in for a fact, it has to be real. If it's standing in for a feeling or an idea, AI can be one of the tools, as long as we're honest about it when it matters. That's not a legal standard. It's a design standard, and I think the profession needs to hold it before somebody else decides to hold it for us.

There's also the uncomfortable question of where these models learned to draw. Many were trained on enormous amounts of work by artists and photographers who were never asked and never paid. Reasonable people disagree on how that should shake out, and the courts are still working on it. I don't think designers can solve it alone, but we can at least prefer tools that are transparent about their training data, and we can keep paying the illustrators and photographers whose work we actually want.

If you're early in your career, my first piece of advice is to learn the tools, all of them, and don't be precious about it. The designers who refused to learn Quark, and later InDesign, didn't preserve the craft. They just got left behind by people who had the craft and the software. Be the person in the studio who knows exactly what Generative Fill can and can't do, and who can get a clean result out of it in five minutes instead of fifty.

But don't confuse fluency with skill. Spend at least as much time on the things the machine can't hand you: typography, hierarchy, grids, color, and above all, the habit of asking why. Why this headline? Why this photo? Why would anyone stop on this page? Sketch by hand. Study the work you admire and figure out what makes it go. Find someone who will tell you the truth about your work, and listen even when it stings. The apprenticeship may not come to you the way it came to me, so you may have to go looking for it.

And build a point of view. When anybody can generate a competent layout, competent is worth less. Your perspective, the thing you notice that other people walk past, is what nobody can prompt their way into.

If you're the one doing the hiring, I'd ask you to think twice before cutting the junior roles because the software can handle the production. It can. But your senior designers didn't become senior by being handed finished work. Keep a seat at the table for someone learning. Pair them with the tools and with an experienced eye. In five years, you'll be glad you did, because that's who will be catching the mistakes the machine makes.

And please, budget for thinking. The cost of making a thing has fallen through the floor. The cost of knowing which thing to make hasn't. If anything, it's gone up, because now there are a thousand plausible options on the table instead of three, and somebody has to have the judgment to pick the right one and the nerve to throw the rest away.

I keep coming back to that email. The one with the sunset, the almost-logo, and the line that says, "Can we just use this?"

I've learned not to get defensive about it. Usually the right answer is a question: "What do you want someone to feel when they see this?" Nine times out of ten, that's where the real conversation starts, and it's a conversation the client couldn't have had with the software. The generated image didn't replace the designer. It just showed up early to the meeting.

AI is going to change this profession. It already has. Some work is gone and isn't coming back, and some good people are going to have a hard time because of it. I don't want to pretend otherwise. But the core of the job was never pushing pixels. It was listening, deciding, editing, and standing behind the result. That's still ours, and I think it's going to be worth more, not less, in a world drowning in competent, forgettable pictures.

So learn the new tools. Use them where they help. Be honest about where they don't. And keep doing the part only a person can do, which is caring enough about the people on the other end of the page to get it right.

The machine can make a picture. It still takes a designer to make it matter.

Will AI Replace Graphic Designers? The Fear vs. The Facts, Clutch

Graphic Designers, Occupational Outlook Handbook, U.S. Bureau of Labor Statistics

Adobe adds its AI assistant to Premiere, Illustrator, and InDesign, TechCrunch

Welcome to the entry-level void, Carly Ayres, It's Nice That

Copyrightability of AI Outputs: U.S. Copyright Office Analyzes Human Authorship Requirement, Jones Day