This article is part of Beyond the Prompt — a campaign exploring Creativity and AI through the voices of industry leaders. Discover more stories, educational content, and our Shortest AI Film competition on the full campaign page.

In my creative work as an Executive Creative Director at Mayda, I’ve always chased after that moment of surprise — the magic that happens when something unexpected slips into the process. Whether through a fortunate accident, a collaborator’s fresh take, or simply messing with a tool until it breaks, that unpredictability has always felt essential to my practice.

That’s probably why AI feels so natural for me to work with. Not because it makes things easier or faster – although it definitely does – but because it opens up new corners of imagination I wouldn’t have stumbled into on my own. Over the past couple of years, I’ve integrated AI into my workflow as both a technical tool and a collaborator. Honestly, I wouldn’t have made some of my recent projects without it.

Look Development with AI for Zedd & Remy Wolf’s Music Video

When we at Mayda directed a music video for Zedd and Remy Wolf last year, AI helped me communicate abstract concepts to a big team. The video’s concept was about the environment reacting to music, with warped spaces and impossible camera moves. Normally, you’d spend weeks with concept artists to visualize something like that – simply to use it as a pitch.

Instead, I mocked up quick AI-generated environments and comped in images from our LA shoot to test how these surreal spaces could feel. The AI gave us a shorthand for talking about impossible ideas and a fast way to visualize the concept of the video, which saved time, money, and frustration. The finished piece was built with traditional visual effects and ended up remarkably close to those early AI-generated images.

For the title sequence of the ‘Lucky’ music video, I illustrated the title by hand, ran it through AI prompts to turn it into shimmering fabric, and then animated it. While legal restrictions meant we couldn’t use AI-generated assets in the final cut, the process helped us test ideas faster than ever. It became an R&D tool – a way to see what works and what doesn’t without burning through days of production time.

‘Bad Genes’: Blobby Creatures and Genetic Experiments

I also started experimenting with AI as a renderer for my own illustration work. In a personal project called ‘Bad Genes’, I took hand-drawn animations and ran them through style models to generate blobby, mutant body forms. I wasn’t trying to control the outcomes too tightly. Instead, I’d prompt AI with a mixture of my images and vague instructions, letting it misinterpret things in surprising ways. I’d then draw over those results and feed them back into the system. Eventually, I trained a custom model on thousands of these images, which gave me a library of weird, grotesque characters that still carried my aesthetic but had evolved into something new. It became a kind of creative ping-pong match between me and the machine.

An AI-Generated Motorcycle Cult for Obtran

More recently, I’ve been working on a fully AI-generated music video for a label called Obtran. The concept I developed is a surreal motorcycle cult in a hallucinogenic world where you’re never quite sure what’s real and what isn’t. I wanted the visuals to land in that strange space between photorealism and obvious synthetic imagery.

AI was perfect for this because it doesn’t have to obey reality. I used it to generate dreamy, distorted backdrops and characters, then filmed those visuals off a screen with physical magnifying lenses and warped glass to mess them up even more. That back-and-forth between digital and physical, AI and analog is where the good stuff happens. It stops the AI from feeling generic and lets me reclaim authorship in the process.

Where the Magic Lives

The biggest thing I’ve learned is that AI isn’t about chasing perfection. It’s about creating a space for randomness and weirdness to enter your process. In the same way I’ve always left room for collaborators to surprise me, I treat AI as another voice in the room, one that can throw out ideas I wouldn’t have thought of.

Sure, everyone has access to these tools now, and a lot of AI-generated images are starting to look the same. That’s why I intentionally build nonlinear, experimental workflows, remixing and reprocessing AI outputs to make sure they still feel personal. The value isn’t in the tool itself, it’s in how you misuse it, break it, and bend it to your will.

I wouldn’t have made these projects without AI. But more importantly, I wouldn’t have made them without letting go of control and letting something unexpected in. That’s where the magic lives.

Learn more about Carl Addy here and about Mayda here.