I Tried Creating a GP Office Concept with Gemini

Hello! This is Kawashima, now entering my third year at GP. Time really flies……

Anyway, before we begin, please take a look at the image below.

What do you think?

This is actually the official “GP Office Concept Guide” displayed inside our office. And surprisingly, the entire design was created using Gemini.

Today, I’d like to share the behind-the-scenes story of how it was made!

About “THE CITY”

First, let me briefly introduce the office that became the theme of this project.

“The WE WORK of the event industry.”
“A city where professionals gather beyond titles.”
“A crossroads where people, information, and culture intersect through the power of design.”

“THE CITY” was created as a space that embodies these ideas. The concept was led by our company president, while GP employees also contributed ideas throughout the design process.

Every piece of furniture and every interior detail was carefully considered to create a space where people naturally stop, start conversations, and connect those moments to new actions and ideas.

This project was born from one strong desire:
“We want more people to understand the philosophy behind this office.”

How the AI Design Was Created

This design was not generated in one shot with a single prompt. It was built gradually through repeated processes and adjustments.

The foundation of the design is based on “isometric illustration,” a visual method that depicts objects from an angled bird’s-eye perspective. You’ve probably seen this style before.

The workflow looked like this:

① Photographing Objects Around the Office

First, I walked around the office taking photos of objects that symbolized GP.

We carefully selected items that visually represented the identity and atmosphere of the GP office.

② Turning the Photos into AI Illustrations (While Unifying the Style)

Each individual photo was transformed into an illustration using a prompt format prepared in advance.

I actually created those prompts through conversations with ChatGPT. I didn’t have any strict rules, but personally, I liked separating the roles:

• ChatGPT for generating and refining prompts
• Gemini for image generation

By repeating this process over and over, we mass-produced a huge number of assets (illustration materials).

And honestly, the illustration of Kawamoto CP became incredibly popular within the company.

③ Arranging the Assets One by One in PowerPoint

Finally, all of the assets created in step two were manually arranged one by one inside PowerPoint……!

(As a side note, one of GP’s production policies is that we basically work entirely within PowerPoint rather than using dedicated design tools.)

If you look closely, you can actually notice small differences in illustration style between assets, and some areas where the isometric perspective isn’t perfectly aligned.

Final Thoughts

So, what did you think?

When people hear “AI-generated art,” many probably imagine creating everything with a single prompt. However, with current image-generation technology, producing a highly detailed and complex design like this all at once is still quite difficult — especially for something important like an office concept guide.

That’s why, this time, we expanded the way we approached AI utilization. Instead of trying to minimize effort, we focused on creating a reliable design process, even if it required more manual work.

In fact, even the isometric composition itself was built by manually placing assets one at a time.

The core idea behind this method was surprisingly simple:
“What if we just followed the same process a designer would use?”

The fact that even non-designers can now create work at this level is one of the most exciting parts of living in the AI era.

I highly encourage everyone to try creating something yourself too!

See you next time!