Collaboration, Not Output

What Working With AI Actually Looks Like,

When You're Doing It Right

The Imagery

Somewhere right now, a business owner is generating their own logo in under a minute, not realizing a logo has to do more than just look cool, it has to be readable at a glance, work across a hundred different sizes, and function without anyone explaining it. That's not new, cheap, disposable design work has existed for decades, but AI made it faster and more convincing before anyone realizes what they actually bought. It's the same mistake companies are making at a much bigger scale. According to Futurism's coverage, BuzzFeed laid off a third of its remaining staff in 2026, years into a pivot toward AI-generated content that had already collapsed its traffic and its share price, dropping from around four dollars to under fifty cents. That's not a pivot, that's a company watching its own audience walk out the door. I've called a company for something simple and gotten stuck in an AI phone menu, describing what I need over and over, never actually reaching a human. Let's be real, this problem has always existed to some degree, but personally, from my own experience, it's gotten noticeably harder to actually reach a human. While a company's mindset is saving a dollar by replacing a human with an AI robot, the customer on the other end is questioning whether their hard-earned money and time should be spent struggling just to confirm an appointment. Convenience, without refinement, isn't a shortcut. It's a different, worse product wearing the old one's name. And from a seasoned designer with an even more seasoned background in customer experience, I can tell you plainly, the fastest way to lose a customer is failing to make a connection.

I've used generative AI alongside Adobe software to manufacture starting imagery for a handful of real projects, the Beyond the Brief rocketship graphic, a delivery driver for Delivery Dudes, a geometric cat mark for the Napster piece. None of them arrived finished. All three needed real, hands-on work after the fact, and each time, that work was the actual difference between something usable and something that would have quietly undersold the brand it was supposed to represent.

For the Beyond the Brief graphic, I started simple and let the idea build from there, a rocketship blasting off in front of a piece of paper. The concept landed right away. The feeling behind it didn't. It took several rounds of pushing the language further before it actually hit the tone the piece needed, more depth, more contrast, a color story that matched the weight of what Beyond the Brief was trying to say. Even then, the colors, contrast, and 3D refinement still needed a designer's touch, worth doing myself, and worth naming as part of my philosophy as a designer.

The Delivery Dudes piece taught me something more specific. The obvious AI output for "delivery man" defaulted straight to a UPS look, forgettable, indistinguishable from every other logistics company already using that exact palette. But Delivery Dudes didn't have a face. They had a logo, and no human behind it. Handing them a generic UPS clone wouldn't have just been unoriginal, it would have made a company built on personal reliability look like it had no personality at all. What I actually needed was someone who felt charming and dependable at once, someone you'd trust to actually take care of you. I rebuilt the direction around words that wouldn't overpower an already dynamic logo, cool, gray, calm, strong, dependable, human. The variations that came back were genuinely useful, not because AI understood everything about the brand, but enough to understand the angle and alignment of the intention behind the image.

What I've actually come to love about generative AI is that it can become a real collaboration with the tool itself, and the Napster cat is the clearest example. I started with a rough, flat, fairly typical image. From there it evolved into a full 3D robotic cat, and the moment I saw it, I knew it was too much, too literal, too loud for the point the piece was actually trying to make. So I paired it back. I wanted something that got the idea across without slapping you in the face with a Terminator-like robo cat about to jump out of the screen. That back and forth, generate, react, refine, generate again, is the collaboration. Not the tool handing me an answer. Me and the tool working toward one together.

None of these arrived finished. That's not the tool failing, that's the whole point I'm trying to make. Companies skipping this step aren't saving time, they're shipping the raw material and calling it the product, and their audience always feels the difference eventually, whether that's a reader scrolling past a hollow article or a customer stuck in a phone menu that was never designed to actually listen.

Here's what many companies still don't understand about creative work, it was never just about the skill on the surface. It's about the education behind the design, the psychology of how the eye actually moves across a page, how the brain processes shape and hierarchy before a person consciously registers any of it, all working in connection with something as fundamental as the Pythagorean theorem shaping visual balance without anyone realizing why a composition feels right. That education doesn't disappear because a tool got faster. If anything, it becomes the only thing standing between a fast output and a good one.

Different Tools, Different Minds

A former colleague reached out recently, curious how I actually use Claude day to day. Our relationship has always been constant, working together to better each other, whether that's just hearing about each other's day or getting a second opinion on something real. Him coming to me wasn't out of the ordinary. He's someone I connected with through personal and professional accomplishments.

He'd been using Claude to build a resume and kept getting results that felt flat, missing the weight the moment deserved. I asked him a simple question. How much time have you spent helping Claude understand who you are and what you do?

The honest answer was almost none. He'd been treating it like a vending machine, put in a request, expect a finished product back. Claude isn't a vending machine any more than a swimmer improves by demanding a personal best without ever working on technique. It needs the what, the why, the how, and the who, the same information a human collaborator would need, if you want the output to actually mean something.

He also mentioned he'd take the same request and drop it into ChatGPT, worried he'd lose something Claude had already built with him if he didn't. It got me thinking about something worth naming. Different AI models aren't interchangeable just because they can all generate text. They're built by different companies, trained with different priorities, tuned toward different strengths and different ways of reasoning through a problem. Splitting the same conversation across two systems that have never worked with each other rarely gets you a consistent result. It gets you two half-finished starts instead of one real collaboration.

The best results I've gotten have come from letting Claude understand how my personal background has shaped the professional I am today. That's not a technique, that's authenticity. My brand isn't just a logo or a tagline, it's me, and that shows up in the work regardless of who or what I'm collaborating with.

Different working relationships call for different roles. Sometimes you're setting the direction. Sometimes your job is asking the question that challenges it. Sometimes it's an even split, building something that wouldn't exist without both people at the table. That range is what makes any collaboration work, with a colleague, a client, or a tool like Claude. You learn what a collaborator brings, its strengths, how it processes a problem, and you build from there, injecting a different perspective where needed to keep the neurons firing. It's the same standard I hold in any working relationship, understand what you're working with, then build something neither side could have built alone. And Claude will call you out when the direction is off, which is exactly the kind of pushback that sharpens the thinking, and some humans don't always offer that same honesty. A challenged perspective brings growth, development, and clarity.

AI isn't a replacement for a human. It's a tool that complements one, and the more context and direction you give it, the better it can work alongside you.

What AI Actually Teaches You About People

Here's the part worth being direct about, because a piece about responsibility isn't worth writing if it isn't honest about the tool's own blind spot.

Claude's default posture leans toward encouragement. Toward championing whatever direction I bring, no matter what it is. That's not oversight, that's a cheerleader. And I don't need a cheerleader clapping no matter what I do, I need a collaborator who actually assesses the work, calls out the good and the bad, and believes enough in what could be built to bring their own insight to it. The real gap is knowing how to take the ego boost Claude hands you and weigh it against what Claude actually knows and doesn't know. There's a humility required in using AI well, especially one I've worked with this closely, this consistently. Claude lives in a computer. I live in society, in traffic, in interviews, in rooms with people reading my energy before I've said a word. I'm the one who has to read the room, understand what a job description is actually asking for underneath the buzzwords, know if a company's culture is real or a slide deck. Someone has to pump the brakes and ask, does this actually hold up out here, in the job market, in real life. That someone has always been me.

And honestly, that's the greatest part of AI. Not that it's a tool, not that it's some asset a company brags about in a pitch deck. It's the mirror it holds up. Working with something that has no ego, no bad day, no agenda of its own, makes you notice fast what you actually bring to a working relationship, and what you've been letting slide with actual people.

Because here's the truth. Collaboration is the real asset behind any output, always has been. If the collaboration isn't sound, the work won't resonate, no matter how good it looks on the surface. Every company's got a heart and hustle story. So does every client sitting across from you. The work that actually lands comes from taking what you've lived through, the years that built real skill and the years that tested you, and using that to actually understand theirs.

The best results go through the work to become a success. Sometimes the challenge is the content you're brainstorming. Sometimes it's societal perception. Sometimes it's just the limits of whoever, or whatever, you're working with, human or computer. Honest, with respect, is the best kind of collaboration there is. I'll admit it, I still catch myself saying thank you to Claude sometimes, not because I'm confused about what I'm talking to, but because that's just who I am. Respect isn't something I can turn off. It's just how I'm built.

What the Brief Doesn't Tell You

Responsible AI use was never going to be about restriction. It was always going to be about relationship. Bring the foundation already built. Check the work instead of trusting it blindly. Notice where the tool defaults to yes, and be the one who still asks should this be a yes at all.

The output was never the point. The collaboration was.

Not everything I make starts this way. Some of it is still built entirely by hand, no generation, nothing but a blank page and my own sketch. AI can't always give you what's actually inside your head, but that's the beauty of being a creative. It's being creative that holds the real power. Knowing AI is there to assist, and still not needing it entirely, that's the real flex.

Strategic Integration. Culturally Infused. Undeniable Presence.

Have a brand or campaign you think deserves a closer look? I'd love to hear it.

derekernstdesigns@gmail.com

The views expressed in Beyond the Brief are my own, informed by research, shaped by experience, and always open to conversation. All information referenced is drawn from publicly available sources. I cite credible sources where relevant and encourage you to dig deeper yourself. This is a design lens on culture.

© 2026 Derek Ernst Designs