Still, Code and Theory isn’t trying to replace the entire production ecosystem with Higgsfield.
“AI has not yet reached the stage where it is completely indistinguishable from real-world action and movement, both in terms of physics and human reactions and emotions.” Foster said. Legal risks such as copyright infringement also influence how agencies choose which AI tools to use. Platforms like Adobe Firefly, which are designed to be more secure for commercial use, are often used to develop characters and brand assets early, while tools like Higgsfield are introduced later for creative management and rapid development of campaigns.
A director of WPP’s innovation team, who was not authorized to speak to the media, told ADWEEK that Higgsfield was attracted to a team that did not have the deep expertise of Prompto Engineering. The platform’s style presets take the guesswork out of you and allow you to generate editorial-style visuals (high-contrast lighting, wide cinematic angles) without creating complex prompts. In reality, Higgsfield is often used internally or presented to clients as a conceptual basis for larger campaigns, rather than as a final product, the executive said.
In internal testing, we find that WPP often runs identical prompts across competing systems, and results vary depending on how each model is trained. The question becomes less about which model is considered “best” and more about which model produces the right creative results with the least amount of friction.
Zero production tax
For Mashrabov, that flexibility is key. He’s not interested in the negative effects of AI currently clogging up his feeds, the uncanny valley of repetitive ads that feel like cheap replicas of mid-2000s infomercials. Rather, he’s betting that the next era of advertising will embrace entirely new concepts that are only possible with AI, making it easier to localize and adapt content across demographics and languages.
“We will use AI to reduce production taxes to zero,” Mashravo said.
Looking ahead, Higgsfield has plans to expand into performance marketing and feed performance data back into the content generation process to help guide which variations are worth creating.
In an increasingly crowded AI video market that attracted more than $500 million in new funding in 2025 alone, Mashrabov is betting that marketers don’t need another black box, but rather a system that helps ideas become important fast enough.
