The Truth About AI Production: Lessons from the KGM Brief

KGM AI Car Ad – Project Debrief

We recently completed an exciting brief for a fully AI-generated car ad and spoke with our creative AI lead to understand the process.

Q: What was the Ask and What We Did

“We were given a client brief along with a storyboard created by AI. Our task was to transform this storyboard into a final deliverable. Rather than following the storyboard strictly, we used it as a starting point and refined the narrative. We then created hand-drawn sketches to visualise the shots before moving into AI image generation. This approach allowed us to focus on storytelling and better capture the essence of the campaign.”

Q: What were the Challenges of the Project

“The main challenges were creating authentic UK environments using only reference images. Another challenge was the car assets—the client initially provided a limited number of real images and 3D models because the products were new to market, these did not integrate seamlessly into AI-generated images. Client requests, such as swapping car colours, added further complexity. Additionally, AI-generated image and video creation are separate processes, meaning images often had to be reworked manually for video.”

Q: How Did We Actually Create It

“Our process involved first creating realistic, client-ready images before moving to video. We used software to build car objects as 3D-like AI assets, which could be placed within generated locations. For video, cars were removed from backgrounds and replaced with AI-created car objects, ensuring correct scale, reflections, and movement. AI models have limitations, such as drivers appearing on the wrong side, so we manually corrected details to make scenes realistic. The project required a combination of creative problem-solving, iteration, and manual compositing to achieve the final result.”

Major Learnings

1. Real-world assets accelerate production – Photographs of locations and more of the cars would significantly speed up the workflow and improve authenticity.
2. AI complements, not replaces, traditional production – Prompts alone cannot create perfect outputs; professional post-production and compositing remain essential for high-quality results.
   

Check out the full [Agency blog post] for more insights into this project.