How to add the AI in the process of creation ?

AI has settled into our daily life as a creative studio. Not all at once, not as a trend, gradually, concretely, starting with editorial tasks, then moving into visual creation. Today, it's a real part of our process. Here's how it actually works.

Two distinct uses, one shared standard

We distinguish between two main ways of integrating AI into our work. The first is the editorial foundation: structuring ideas, drafting a brief, testing a communication angle. Claude serves here as a thinking surface, a tool to organize what's in our heads before we even speak to a client.

The second is visual creation. And that's where things get more interesting, and more demanding.

Nano Banana: a concrete cxample

Working with AI doesn't look the way people often imagine from the outside. It's not: write a sentence, get an image, use it. It's a much more iterative process.

Sometimes we start from a blank page. Sometimes we start from a 3D model we've built ourselves. In either case, the base is a starting point, not a result. What follows is work: adjusting materials, working the light, finding the right atmosphere, integrating human elements, refining visuals until the whole thing feels coherent and alive.

Prompting is a craft in itself

There's a skill we had to develop that we underestimated at first: knowing how to prompt. Writing an instruction for an AI tool is its own form of creative writing. Too vague, and the result is generic. Too rigid, and you constrain what the tool can bring. You have to find the right level of constraint, know when to allow room for interpretation, and above all know how to read what you get in order to course-correct.

And after the prompt, there's the retouching. That's often the longest part. AI produces raw material that we shape: we correct, replace, layer, recompose. The creative eye remains ours, AI accelerates certain steps, it doesn't eliminate any of them.

What it changes (and what It doesn't)

Integrating AI doesn't mean producing faster by doing less. It means exploring more broadly with the same high standards. We can test visual directions we wouldn't have had time to prototype otherwise. We can iterate on a lighting mood in minutes rather than hours.

But the final quality always depends on the same factor: the eye. Knowing what works, what feels off, what's missing. AI doesn't replace that judgment. It gives us more material to exercise it.

And sometimes it feels wrong...