Case Study: HARMAN drive
Generative AI, design, and strategic storytelling combine to deliver stories of team success.
Overview
To help HARMAN Automotive bring its new cultural values to life, we created an internal comic book universe powered by generative AI, bold design systems, and strategic storytelling. Each issue of the How We Did It series transformed real employee wins into cinematic stories of teamwork, innovation, and grit, casting HARMAN’s six core values as superpowers.
Adopted globally, featured at CES, and even embraced with cosplay and fan art in multiple countries, the campaign quickly became more than communications, it became cultural canon.
The opportunity
As HARMAN transitioned from tier-one supplier to tech-first innovator, leadership sought a fresh, engaging way to energize employees and unify the company’s evolving identity. The challenge: make internal values tangible, memorable, and motivating for thousands of employees across global teams.
The strategy
As creative lead, I partnered with account and strategy teams to create six HARMAN values, branded as superpowers, each with its own emblem, color system, and character identity. We developed a visual and narrative world that struck a balance between modern graphic novel and retro comic, built on HARMAN brand assets.
From story sourcing to design to final layout, I led the creation of every comic volume, combining character-driven narratives, generative AI visual exploration, and custom design to deliver employee-focused stories that honored the real heroes: HARMAN teams.
In my role, I designed and produced:
- Superpower identities and DRIVE hero characters
- Storyboarding and visual language for each issue
- PanelPilot, a Custom GPT-assisted character generation tool
- Cross-channel rollout materials and in-room visuals
- Narrative strategy to support internal adoption and sustained engagement
- Final comic book layout and design for each issue
Technologies & tools
- ChatGPT
- Dall-e
- Midjourney
- Figma
- Photoshop
- Adobe Animate
- iPad Sketchbook app
PanelPilot: Custom GPT-Powered Comic Rendering Engine
To scale the comic series while maintaining consistency, I built a custom GPT, an instruction-driven, stylized visual engine capable of generating high-fidelity, cel-shaded scenes on demand. The tool translates structured prompts into cinematic comic art using a predefined cast of HARMAN heroes, environments, and narrative styles.
Results
Seven full volumes (and counting) have been produced to date, each celebrating real-world business wins and teamwork across geographies. The impact has been both measurable and cultural:
- 96% employee awareness of the superpowers
- 80%+ see them reflected in daily work
- 70% say their leaders actively support the values
- Comic art and characters recreated across global offices—including cardboard heroes in China, live-action appearances in Mexico, and hallway murals in Germany
- Prominently featured at CES 2024, with new issues integrated into HARMAN’s global engagement strategy
The series became a pillar of HARMAN’s internal brand, helping drive culture change from the inside out, with more issues, spinoffs, and story arcs still to come.
Project gallery
Early concepts
Initial Explorations: Blending AI, design and comic panel design
To create the sense of a true comic book experience it had to feel like a comic in style, flow and structure. Dynamic, dramatic, breaking the panels so the story comes to life in mood, texture and atmosphere. As person who is a traditional illustrator, it was particularly fun to do but my skills in illustration were also essential in the overall exercise.
Exploded View
This just shows the process and thinking behind each page. We didn't go in the direction of this visual style, but this allowed our client to really see what was possible.
Settling in on a style
Issue #7 of 7
My illustrated character art and generated background imagery.
Each panel is a combination of generated character portraiture, design and layout.
Here is a breakdown of the individual elements of the the page above. I used PanelPilot to render story characters and touched each one up individually when needed then composited them into each panel layout.
Issue pages continued.
From screen to stage.
Office Wall Mural at Harman HQ. The client's reaction is priceless.
Planned Collectibles
Creating PanelPilot
The start of the experiment. Teaching and building.
Team shot: Which I traditionally illustrated and delivered as vector files.
Colors were sampled and labeled to determine cel-shading light and shadow
I instructed GPT to use these character design sheets along with my preferred output structure to generate consistent description files.
I then named, provided a description, created instructions and an initial conversation starter to 'prime' the GPT.
It was done and now it was time to test.
PanelPilot results
To say that I was satisfied is an understatement. By entering just a character name and expression as a prompt, i.e. "Trustmeister, Focused", I was able to generate the images you see below.
The final output features over 50+ portraits across multiple characters and mood states, each fully on-model and stylistically unified. The client was thrilled with both the visual fidelity and the speed of turnaround. What would have taken weeks of manual work was delivered in days.
Takeaway
This project validated the effectiveness of a prompt-driven visual system for scalable character art production, blending consistency, speed, and creative control without compromising quality.