For most of my career, learning creative tools followed a somewhat predictable path. The more effort I put in, the more I was able to not only improve my skills, but the more I started to grasp the underlying concepts.
I studied copywriting at an advertising academy and later taught myself 3D design, spending more than two decades working with complex tools and production pipelines. Whether it was writing, Photoshop, Blender, or Cinema 4D, the learning curve always felt difficult yet logical.
You started with clumsy results, slowly developed technical control and eventually gained something far more valuable than speed. You were able to offer taste, judgment, and visual literacy.
Then AI arrived.
Besides all the ethical implications and dissolution of the workforce etc.; designers, creatives and visual thinkers have been handed another unforeseen symptom.
It seems that the learning curve has completely flipped.
Instead of struggling towards quality, designers can now start almost immediately with seemingly impressive results. This is not only true with image generation or animation, but also with writing, coding and particularly with research. The initial results are high above the novice level, giving any user an almost immediate boost to their skills, but paradoxically this makes learning harder, not easier.
This is what I believe is one of the core challenges of using gen AI in design: visual quality now comes before understanding.
The Traditional Design Learning Curve
In my experience, the more traditional the design tool, the more it forces you to earn your results. We could assume something like sketching with a pencil and paper is on the side of the most traditional and all the way on the other end of the graph is Gen AI.
This phenomenon isn’t exactly new either since tools like 3D modeling and rendering already cut down the perceived need to learn the foundations. Once you jump into 3D software, you’re immediately presented with some tools that can “render” your image with quality that was previously unattainable when using “traditional” methods like pen, paper and copic markers.
One of the earliest modern examples of this might have been photography, revolutionizing the user's ability to represent a subject realistically. But even photography required foundational skills such as an understanding of lighting and composition in order to render a compelling image.
Starting at photography, with each jump in technology, the user needed less and less foundational skills to render a baseline image. The difference with Gen AI is that baseline image also includes many of the foundational skills, allowing the user to not only generate a compelling image, but one that includes some idea of composition and lighting that they may have not considered, or happened by accident.
So traditionally, as you learnt new hard skills, you developed multiple layers of softer skills at the same time such as compositional awareness (balance, contrast, rhythm, hierarchy), conceptual thinking (meaning, storytelling, communication) and aesthetic judgment (what feels right and why).
Every mistake was teaching you something and very obvious limitations forced you to understand the mechanics behind each decision.
With this paradigm, the relationship was fairly linear (even though it could feel like the more you uncovered the less you knew).
More skill → deeper understanding → better outcomes.
The AI Learning Curve: Instant Output, Delayed Understanding
Now more than ever, the new tools at our disposal are disrupting this process of learning. With tools like MidJourney, Nano-Banana or any frontier diffusion model, you can produce cinematic visuals, polished layouts, or complex compositions in minutes without formal training.
This isn’t an argument over whether that’s a good or bad thing, it’s simply an observation in how our ability to generate outputs have evolved.
But due to this perceived ease of creating visual results that look and feel like the real thing, we risk losing our ability to judge for ourselves whether an image is better or worse than another one. Or if it’s not worse or better, then how its messaging might differ.
Now we have this strange gap where the output looks professional but the understanding behind it might be missing. You might generate an image that feels balanced, dramatic, or emotionally resonant, but struggle to explain why the composition works or how the visual hierarchy guides attention.
In traditional design, skill precedes quality. In AI-driven design, quality often precedes skill. This is why learning design with AI feels so disorienting.
Designing Backwards: From Output to Understanding
This all sounds like a problem that might not be relevant as we evolve. Who needs all this understanding when you have an AI that can basically do all the foundational things for you and then you can play Creative Director.
And yes, this doesn’t really matter, at least when you’re still playing around and creating images or visual outputs that don’t require any specificity. It’s much easier to create outputs when the parameters are defined as you go along, than to say work to a brief or some visual guideline which creates a need for specifics.
When you’re creating in this manner, you can choose the image with the best composition and then adapt your expectations to the fact that the subject is no longer exactly what you had in mind when you started. However, when you’re attempting to make something specific, the visual language becomes much more important.
We’ve all been there before. One or two prompts and you’re calling your colleagues over to your desk to marvel at the miracle solution that you’ve been handed. You blink for a second longer than usual and enter a dream of going home at 4:30pm.
Your colleagues pat you on the back but just before they leave, your creative director pops in and points out that you need more space at the bottom for the client’s logo. No problem, you think, I'll just quickly add some prompt to give me more space. You get more space, but then the image becomes cramped.
Suddenly things are starting to get messy. The prompt is getting more complex, birds are suddenly multiplying in the scene and you don’t quite understand where your image is going. In fact you no longer understand where you started and you’re relying quite heavily on AI to solve the design problem. What we fail to notice is that just because AI can easily solve a simple image, it doesn’t quite understand how to fix it once it goes off track, not unless we can give it that direction. And the knowledge of what the image needs to be solved requires a bit of vision, a bit of vision that you have learnt over the years.
It reminds me of the story about the woman who approached Picasso in a restaurant, asked him to scribble something on a napkin, and said she would be happy to pay whatever he felt it was worth. Picasso complied and then said, “That will be $10,000.”
“But you did that in thirty seconds,” the astonished woman replied.
“No,” Picasso said. “It has taken me forty years to do that.”
Now I don’t want to pretend like this is an unsolvable problem, and quotes about what AI can’t do are aging pretty badly. I’m merely illustrating that there is a trap that we need to avoid, and it’s simply thinking that we get past stage two of the AI design process without the hard earned skills. Which brings me to my next point… What exactly happens when we lose the ability to increase our skills?
The Hidden Risk: Losing the Training Ground
Gen AI is improving incredibly fast. So fast in fact that what I'm writing here today might be outdated in a few months or even weeks.
With this pace comes what I consider the biggest tradeoff. Historically, junior designers developed visual literacy through repetition and struggle. They spent years rebuilding compositions, adjusting typography, experimenting with lighting and color - failing, iterating, and refining.
But the catch is that these “inefficient” processes were not wasted time. They were how designers learned to see. AI now automates many of these formative tasks.
And so we have a small paradox, namely that AI replaces the very activities that used to train designers. So if beginners can skip the struggle and jump straight to polished results, where does their visual intuition come from? In a usual design ecosystem they would essentially not become designers, as they would continue to lack the skills and not find employment, or they would find small jobs and continue to improve as shown earlier.
But now their work is validated in the marketplace and they can continue to exist as designers at a higher perceived level. If machines solve the technical problems, who learns the underlying principles?
And so, we risk producing a generation of designers who can generate images but cannot judge them. It’s good to note here that this paradox might not only apply to design, but any task where some training was precipitated through a process of repetition.
The New Designer Skill Stack
So, if AI changes how design is produced, it should also in turn change what designers must focus on when they are going through this process.
Technical execution is no longer the primary differentiator for designers entering the workplace, instead the most valuable skill becomes visual literacy, critical judgment and taste.
Ironically, these are the skills that were previously developed partly through technical struggle. Now they must be learned deliberately, possibly as a focused and separate task.
Final Thoughts
Let’s be clear, AI does not eliminate the need for designers but it might dramatically change what it means to be one.
The future designer was never going to be defined by technical mastery alone, in fact no good or great designer is. The change is that they had the time and place to master the critical skills to become a good designer.
The true challenge is not whether AI can make beautiful visuals, it becomes whether designers can still learn to understand them.
And that is why I believe that the AI learning curve does not rise step by step, instead it runs backwards. And while I may not offer any immediate or concrete solutions in this post, sometimes the recognition of a problem is the first step in finding the solutions necessary.
We are already critically assessing the role in AI in our studio and where it can create a void of necessary judgement. Perhaps these techniques can be explored in a future post!
