How I Turned a Problem Into an Opportunity

Creating with no limits, while on a budget. Here’s what worked for me—no theory, just the steps that paid off.

I didn’t start scemax because I woke up one morning and decided the world needed another creative studio.

It started because I had a problem.

I was working on a large commercial project with more than 50 products that needed to be photographed and turned into a full visual campaign. We had a good budget. The problem was that a “good budget” doesn’t stretch nearly as far as it used to.

Once you start adding locations, sets, models, props, photographers, lighting, travel, styling, retouching, permits, insurance and all the other little things that come with a traditional production, the number gets big very quickly.

I had two obvious choices: shrink the creative idea or shrink the profit.

I didn’t particularly like either one.

So I started looking for a third option.

1. I stopped asking what we could afford to shoot

Instead, I started with the image I actually wanted to create.

That sounds simple, but it completely changed the process.

Rather than designing concepts around the limitations of a physical production, I designed the campaign first and then figured out which parts actually needed to exist in the real world.

Some things needed to be photographed. Some didn’t.

That distinction turned out to be pretty important.

2. I started treating AI like a production tool, not the creative director

AI gave me ways to build environments, test ideas and create things that would have been expensive—or sometimes completely impractical—to produce traditionally.

But it couldn’t decide what the campaign should feel like.

That part still needed a person.

The concept, styling, composition, product accuracy, brand consistency, lighting direction and endless tiny decisions still came from me. AI simply gave me another way to execute them.

That became the foundation of what would eventually become scemax: human creative direction, using new tools to make bigger ideas possible.

3. I mixed methods instead of choosing sides

I stopped thinking in terms of photography versus AI.

The better question was:

What is the smartest way to make this particular image?

Sometimes that meant photography.

Sometimes it meant building the environment digitally.

Sometimes it meant compositing real products into a created scene, retouching, rebuilding elements or combining several methods until everything looked like it belonged together.

The goal was never to use AI just because we could.

The goal was to get the shot.

4. I spent the budget where people could actually see it

Once we stopped spending money simply because “that’s how productions are done,” we could put more of the budget toward the parts that actually affected the final work.

Better concepts.

More variations.

More environments.

More freedom to experiment.

And importantly, less pressure to settle for something simply because we had already spent too much getting there.

5. I realized the problem wasn’t unique to me

That project worked.

But the bigger realization came afterward.

If I was running into this problem while working with established brands and legitimate commercial budgets, smaller businesses had to be feeling it even more.

There are a lot of businesses with really good products, really good ideas and really ambitious visions that simply don’t have traditional campaign money.

That didn’t mean their ideas needed to be smaller.

It meant the production model needed to change.

And that’s where scemax came from.

What started as me trying to solve one creative problem became a completely different way of thinking about production.

Not how little can we spend?

Not how much can AI replace?

But:

How much more can we create with what we already have?

Turns out, quite a lot.

And we’re still finding out just how far we can push it.

That question eventually led me into a much deeper experiment: how far can AI actually go when the standard isn’t “good for AI,” but genuinely good creative work?

I started testing where it holds up, where it falls apart, and whether it can really perform at the level of execution I’m aiming for.