AI Couldn’t Count to 20 — So We Worked Together

AI had great ideas. I just had to teach it how to count to 20.

Recently, a client asked me to design a commemorative jewelry pin featuring ivy, diamonds, gold strands, and exactly 20 pearls.

That last requirement mattered.

Not approximately twenty.

Not “about twenty.”

Exactly twenty.

Like many designers today, I turned to AI to help explore the concept. Within minutes, it produced elegant gold vines, luminous pearls, sparkling diamonds, and intricate ivy leaves. It generated idea after idea, each one visually beautiful.

There was just one problem.

It couldn’t count to 20.

Every version came back with a different number of pearls. Twenty-three. Eighteen. Twenty-two. Sometimes pearls merged together. Sometimes they disappeared behind leaves.

The more I asked for exactly twenty, the more I realized something fascinating.

AI wasn’t really following the production specification. It was trying to make the design look aesthetically pleasing. It was creating what seemed visually balanced, even when that balance contradicted the client’s instructions.

AI was designing what looked right.

I needed to design what was right.

Then something happened that confirmed what I was seeing.

The client took the concept to a jeweler.

The jeweler was also using AI.

And the client came back to me with the exact same request:

“Can you create this design with only 20 pearls?”

That was my “aha” moment.

I stopped asking AI to create the finished pin.

Instead, I asked it to make the individual ingredients:

One pearl.

Three different sizes of pearls.

Gold ivy leaves.

Diamonds.

Curved gold strands.

Then I opened Photoshop.

I counted every pearl.

I resized them.

I moved every leaf.

I connected every strand.

I art directed the composition until it matched the brief.

Together, we got there.

AI supplied the raw materials. I supplied the judgment, precision, and production thinking.

And that is when I realized this wasn’t really a story about AI failing.

It was a story about learning how to use a new tool.

Learning a New Language

Using AI often feels like learning a new language.

At first, you think you are being perfectly clear. Then the tool responds with something completely different from what you intended. So you change the wording. You add context. You simplify the request. You divide one complicated task into several smaller ones.

Eventually, you stop trying to force the tool to think exactly as you do.

You begin learning how it works.

That process reminds me of when I first learned to use a Macintosh.

The mouse felt completely unnatural. My eyes, brain, and hand had to learn to work together in an entirely new way. When the mouse was positioned upside down, I had to retrain myself so that up was down, down was up, left was right, and right was left.

It felt like a puzzle.

But that was also what made it exciting.

Eventually, I stopped consciously thinking about the mouse. The movement became instinctive. The tool disappeared, and I could concentrate on what I was creating.

The same thing happened with Photoshop.

Layers.

Masks.

The Pen tool.

Bezier curves.

At first, every new function required thought. Then, over time, they became part of my creative vocabulary.

AI feels much the same.

At first, you translate every idea into instructions.

Then you begin to understand how the tool interprets those instructions.

Eventually, you stop asking:

“How do I make AI do this?”

And begin asking:

“What is the best way to solve this problem?”

Sometimes the answer is AI.

Sometimes it is Photoshop.

Sometimes it is Illustrator.

Sometimes it is a pencil.

And sometimes it is all of them working together.

The Kitchen-Sink Stage

This also reminds me of the early days of website design.

Suddenly, people could use fifteen fonts, blinking text, spinning logos, animated GIFs, visitor counters, scrolling messages, and music that began playing the moment a page opened.

Because the technology made all of those things possible, people assumed they should use all of them. The entire kitchen sink went onto the website.

Eventually, we learned that good design is not about using everything a tool can do.

It is about restraint.

Purpose.

Hierarchy.

Knowing what to leave out.

AI is going through a similar stage now. I t can generate endless possibilities, and the temptation is to accept the most elaborate or visually impressive result simply because the tool produced it.

But possibility is not the same as intention.

Beautiful is not always the same as correct.

And “close enough” is not the same as exactly 20 pearls.

Don’t Fight the Tool. Learn the Tool.

A master carpenter does not blame a hammer because it will not turn a screw.

They choose the right tool.

The same principle applies here.

I could have continued asking AI to regenerate the entire pin, hoping that eventually it would count correctly.

Instead, I changed my workflow.

I learned what the tool did brilliantly.

I learned what it struggled with.

Then I bent the process to get the outcome I needed.

That is what creatives have always done.

We solve problems.

The tools change—from markers and X-Acto knives to Macintosh computers, Photoshop, Illustrator, digital photography, and now AI.

But the creative thinking remains.

And perhaps that is why I find AI so fascinating. It does not necessarily make me think less.

It makes me think differently.

It gives me a new puzzle.

The people who get the most from AI will not be the ones who expect it to do all the work. They will be the ones curious enough to keep asking:

“What happens if I try this?”

Then:

“What if I phrase it differently?”

And finally:

“What if I completely change the workflow?”

That is not simply automation.

That is creativity.

AI imagined the pin.

I made it accurate.

AI generated the possibilities.

I directed the final solution.

The finished design was not created by AI alone.

It was not created without AI, either.

It was created through collaboration — by understanding what the tool could do, recognizing what it could not do, and finding another route to the answer.

Creative people have never been defined by their tools. They have always been defined by how they learn to use them.

And the future does not belong to the people with the newest tools.

It belongs to the people who learn how to use them.