Using AI as a Creative Tool for Building Custom 3D Brand Worlds

Context

Bestie.work is a talent and referral network built around a more human way of matching people with opportunities. Rather than focusing only on resumes, job titles, or traditional hiring signals, the platform is designed to help people find roles and communities that are better aligned with their personality, relationships, and lived experience.

That idea created an interesting brand challenge. Most hiring and recruiting platforms rely on a familiar visual language: clean dashboards, professional icons, abstract people illustrations, and neutral SaaS-style graphics. Those conventions can make a product feel credible, but they can also make it feel generic. For Bestie.work, the goal was to create something more memorable, more emotional, and more closely connected to the idea of discovery.

The brand direction developed around the metaphor of exploration. Instead of visualizing hiring as a pipeline, funnel, or database, we treated it more like a journey through a series of new worlds. The core visual concept involved a tall ship moving through open water, discovering small islands, remote villages, landmarks, and communities along the way. It created a more playful and narrative-driven way to talk about finding the right role or the right professional community.

This direction allowed the brand to move away from expected hiring imagery while still staying connected to the product’s purpose. The visuals suggest curiosity, movement, and possibility — all useful ideas for a company helping people discover where they fit.

Screenshot of the end-client user interface showing a simplified AI experience.

The Visual Challenge

Once the creative direction was established, the main challenge became execution.

The visual system relied on a very specific style: soft 3D isometric environments, Phong-style lighting, whimsical landscapes, and small self-contained scenes that could sit in the background of a marketing website. The images needed to feel detailed enough to be distinctive, but controlled enough to work as part of a larger brand system.

Traditionally, producing this kind of imagery would require a full 3D workflow. Each island, building, ship, lighthouse, or landscape element would need to be modeled, textured, lit, rendered, and then revised. That approach would provide a high level of control, but it would also introduce significant time and cost — especially for a startup brand that needed a full set of website graphics, not just one hero image.

The project needed a faster way to create a cohesive visual world without sacrificing composition, quality, or brand consistency.

For a less AI-native marketing team, the blank prompt box creates too much responsibility. It asks the user to know what context to provide, how to phrase the request, how to structure the output, and how to guide the model toward the right tone. Even if the team has brand guidelines, competitive analysis, and strong examples, those materials do not automatically make their way into the prompt.

That meant the tool needed to do more of the thinking in the background.

The goal was to create an interface that felt simple on the surface, while quietly packaging the right context, prompts, and reference material behind the scenes. The user should not need to think about prompt engineering. They should only need to answer a much simpler question: What would you like to create?

Approach & Strategy

The first phase of the process focused on visual exploration. We used Midjourney to generate early reference images for the world: small island environments, nautical details, village-like structures, and soft 3D landscape forms. These images were not treated as final deliverables. Instead, they became visual anchors that helped define the tone, lighting, material quality, and overall atmosphere of the brand world.

This was an important distinction. The goal was not to ask an AI model to “make the brand.” The goal was to use image generation as a fast way to explore possibilities and identify a repeatable visual direction.

Early outputs helped confirm that the world could work. They captured the right sense of scale, softness, and whimsy. But they also exposed the limitation of relying too heavily on generated images. The results were visually interesting, but they were static. They did not always fit the actual layout needs of the website, and they were difficult to extend with precision.

A generated image might have the right feeling, but the wrong composition. Or it might include the right type of island, but place the focal point in an unusable area. For web design, those details matter. Background images need to work around headlines, buttons, responsive breakpoints, and scrolling sections. The image model could create beautiful scenes, but it could not reliably design within those constraints.

Where Generation Fell Short

After the initial exploration, we tested ways to push the generated style further. The idea was to create more variations from the existing references and see whether the model could produce a broader set of usable scenes. This included experimenting with reference-based generation and additional prompting to create new landscapes in the same style.

The results were helpful, but not controlled enough.

The model could produce “more of the same,” but it struggled to create the exact scenes needed for the website. When a specific composition was required — for example, an island with a lighthouse in a certain position, or a landscape that needed open space for copy — the generated outputs became less useful. They looked like finished images, but they were not necessarily functional design assets.

That became the turning point in the process. Instead of continuing to ask the AI model for complete images, we shifted the workflow back into a more traditional design environment.

A Hybrid Production Workflow

That became the turning point in the process. Instead of continuing to ask the AI model for complete images, we shifted the workflow back into a more traditional design environment.

This changed the role of AI in the workflow. It stopped being the tool that decided the image and became more like a production assistant inside a larger creative process.

The workflow looked roughly like this:

Generated reference → Photoshop canvas → Manual composition → AI-assisted additions → Color, lighting, and cleanup → Final website graphic

Within Photoshop, we could copy, mask, paint, clip, extend, and rearrange elements by hand. Then, using generative tools inside the canvas, we could create new details based on the surrounding visual context. For example, instead of modeling a lighthouse from scratch, we could define the rough space where the lighthouse should appear, describe the desired element, generate a few options, and then manually refine the best result.

This process gave us the speed benefits of AI while preserving the designer’s control over composition, hierarchy, and final polish.

Building the Brand World

Using this hybrid method, we built a set of custom landscapes and island scenes for the Bestie.work website. Each image was assembled with a mix of generated material, manual editing, and AI-assisted additions. The process was iterative: create one section, extend the canvas, add new landforms, introduce structures, adjust the lighting, blend the edges, and repeat.

Because everything was assembled in a shared visual environment, we could maintain consistency across the system. Lighting, perspective, color, and texture could be adjusted globally. Elements that came from different generations could be blended into a single world. The final graphics felt cohesive because they were composed and finished by hand, even though AI helped create much of the raw visual material.

This was especially useful for a brand like Bestie.work, where the imagery needed to feel expansive. The website could not rely on one hero illustration alone. It needed a repeatable visual system that could support multiple sections, backgrounds, and storytelling moments.

Solution

The final visual approach gave Bestie.work a custom 3D brand world without requiring a full 3D production pipeline. The graphics support the company’s larger brand metaphor: helping people explore new opportunities, discover communities, and find places where they belong.

The imagery feels intentionally different from typical recruiting software. Instead of showing resumes, avatars, or generic hiring icons, the website uses landscapes, ships, islands, and small architectural details to create a sense of movement and discovery.

The result is a visual system that is:

• Distinctive enough to separate Bestie.work from conventional hiring platforms

• Flexible enough to support multiple website sections

• Cohesive enough to feel like one branded world

• Efficient enough to produce without modeling every asset from scratch

• Controlled enough to meet real layout and composition needs

Impact

The biggest benefit of this process was not simply speed. It was the combination of speed and control.

A fully manual 3D workflow would have given us control, but at a much higher production cost. A fully AI-generated workflow would have been faster, but too unpredictable for a polished marketing website. The hybrid approach created a better middle ground.

For the client, this meant BrandZap could create a rich, custom visual system without the budget or timeline usually associated with fully modeled 3D environments. For the design process, it showed how AI tools can meaningfully improve production when they are used inside a controlled creative workflow.

The designer still made the core decisions: composition, visual hierarchy, editing, cleanup, color, and final direction. AI helped reduce the production lift of creating individual visual elements, but it did not replace the need for judgment.

AI image tools are often framed as replacements for creative work. In practice, they are often more useful as collaborators inside the creative process.

For Bestie.work, AI was not used to generate a final image and call it done. It was used more like a brush, a texture source, or a rapid production partner. The final visuals came from a hands-on design process that combined generated imagery, manual composition, and careful finishing.

That is where these tools become most valuable: not when they remove the designer from the process, but when they help the designer move faster while still making the decisions that shape the final work.