L’Oréal has tested an approach in Australia and New Zealand in which generative AI does not make the finished advertisement but challenges the team during strategy development. For a CeraVe campaign aimed at men, the company worked with Australian startup Springboards and its Flint system. Forty creative directions were produced in two weeks; people then refined the chosen route through repeated prompting, discussion and local review.
The problem behind the experiment
CeraVe had not previously built a separate male campaign in the region, so there was little local history to copy. The task was to find both an audience and a social context capable of stopping the scroll without feeling imported. Springboards combined the brief with panel and Nielsen insights, brand guidelines and product information. As launch coincided with football and rugby playoffs, sport became one route into the audience.
- 40 directions were developed in two weeks.
- Australia and New Zealand were the test markets.
- Social platforms formed the main campaign environment.
- Human marketers edited and approved the final idea.
Selection matters more than generation
Generative brainstorming often creates many safe variations of the same thought. Flint increases variability around selected “critical” points, but novelty is not quality. Every suggestion still needs to be checked against the product, category rules and culture. The final line, “CeraVe, Your Skin Barrier’s MVP,” evolved from machine suggestions but was reviewed and rewritten by the team.
What should be measured
Forty ideas and a faster process measure productivity, not campaign effectiveness. L’Oréal still needs to test whether men paid more attention, understood the product benefit and responded better to local creative than to a universal execution. It should also examine whether a sports frame excludes men who do not identify with that interest.
UA Banner has discussed the difference between media scale and business effect. In this case, AI-assisted concepts should be compared with control creative, with brand memory and purchase intent measured alongside clicks.
UA Banner conclusion
This model is more mature than the promise that “AI will make the campaign.” The system expands the option space, the brand supplies verified context and people remain accountable. The real proof will not be the number of generated directions but whether the chosen one reached a new audience without reducing men to a sports stereotype.
Sources: AdExchanger and L’Oréal information about its regional Beauty Tech programme.
