Mazda's CX-50 Campaign Was Built by AI

Mazda chose the CX-50 launch to test something different. Instead of a traditional production pipeline — agency briefs, directors, crews, location scouts — they handed the project to an AI-driven workflow managed by Luma AI.

The tool handled the full project. It generated, iterated, and refined the output on its own until it reached the final cut. No reshoots. No waiting on weather. No day rates.

Publicis and Serviceplan were the agencies on the account. They didn't just use the tool as a finishing trick on top of live footage. They used it to produce the spot.

Publicis is not a boutique. It's one of the biggest agency groups on the planet. When a network that size commits to an AI production model on a real client campaign — not a proof of concept, not an internal test — that's a signal.

Serviceplan has been pushing in this direction too. The combination of both agencies on the same job suggests this wasn't an accident or a stunt. Someone at Mazda signed off on this knowing exactly what they were doing.

The CX-50 is a real product launch. Real media spend behind it. Real brand risk. They still went with it.

That's the part worth paying attention to.

Luma AI generates video from text and image prompts. It's built around neural radiance fields and video generation models that can produce photorealistic footage without a camera ever being pointed at anything.

For a car campaign, that matters a lot. Car commercials are expensive precisely because of the physical production — the locations, the car rigs, the lighting setups, the permits. A tool that can generate a car driving through a landscape, lit correctly, at the right time of day, without any of that infrastructure is not a small thing.

You can read more about why Luma has been deliberately positioning itself close to the creative industry in this piece. They're not trying to replace agencies from the outside. They're getting inside the workflow first.

This is where it gets contentious.

When Publicis used this model, the reaction from the creative sector was immediate. Controversy is the polite word for it. People are angry. Directors, DoPs, production designers, colorists — everyone who gets hired when a car brand shoots a campaign is looking at this and doing the math.

And the math is uncomfortable.

I'm not going to tell you AI production is fine and everyone will find new work. I don't know that. Nobody does. What I do know is that a client already approved this, an agency already delivered it, and the campaign is out.

The experiment is over. The result exists. Now we find out if it's a one-off or the beginning of something that scales across every mid-budget automotive campaign in Europe.

The video ends with a question: ¿Quién será el siguiente? Who's next?

It's the right question. Not because it's dramatic, but because the decision has already been de-risked once. Mazda took the first-mover risk. The next brand that does this isn't pioneering anything — they're following a precedent.

That's how these things spread. Not through one big announcement. Through the second and third adoption, when it stops being news and starts being a line item in a production brief.

If you work in production, in creative, in content — the question isn't whether this will affect your market. It's whether the next job you pitch against will be priced by a human team or by a model that charges per generation.