Ask Fiona Menzies whether Eden House uses AI in its marketing and the answer isn’t yes or no, it’s a spectrum. This year her team made roughly a dozen brand films for different property developments, and they range across the entire scale: one end fully live-action with real casting, sets, lighting crews and a real budget; the other end “fully AI, cast, wardrobe, style, locations, lighting, everything that has been completely 100% done with AI.” Most sit somewhere in between, a live photoshoot with AI-generated animation layered on top, or the reverse.
The honest verdict, from someone who’s actually run both processes back to back, punctures a lot of the industry’s cost-saving pitch. “Budget-wise, it’s not that much cheaper,” she said. “Trying to make that last film that’s fully AI look real took many more hours than it took to edit the real footage of real people.” And the thing AI still can’t manufacture, in her experience, is the reason anyone makes a brand film in the first place. “It’s very hard to create like a spark or a connection or goosebumps with AI,” she said. “I just don’t think it really works. Probably like writing, nothing surprises you. There’s no style or flair.”
Brands Choices
That’s one brand’s answer: use it selectively, expect efficiency gains that don’t always show up, and don’t expect it to do the emotional heavy lifting. Yohan Wadia, founder of Wadia Studio, sees the same selectivity from the agency side of that relationship, built on what he calls “reckless imagination, unapologetic GenAI fire.” His studio runs entire production pipelines on generative tools, but he’s clear that brands aren’t handing over the emotional or reputational stakes along with the workload. “Brands are already cautious, and honestly, they should be,” he said. “But they’re not usually worried about one platform having a bad week. Their concerns are much closer to the work: Is it legally safe? Will it look cheap? Could it damage the brand? Is someone actually in control of the output?” The imagination can run wild, in his account, but the process around it can’t. “There are human decisions at every important stage, from the initial concept and visual language to character consistency and final delivery,” he said. Trust with a brand client, on his telling, is earned the same way Menzies earns it internally: by proving the process, not the platform, is in control. “If a brand can see that you understand its identity, that the output is intentional and that there are proper safeguards in place, the conversation becomes much less about fear of AI and much more about what AI allows them to create.”
Where is the money?
Where brands are demonstrably putting real money behind AI right now is media, not creative. Grubhub has been running ChatGPT ad campaigns since OpenAI opened the platform to advertisers in February 2026, and brought in AppsFlyer’s new Conversion API integration to tie those campaigns back to app installs and in-app orders. “Because so much of our customer engagement happens in the app, app attribution gives us a much more complete view of campaign performance,” said Brian Ryu, Grubhub’s VP of Growth, in the integration’s launch announcement. AppsFlyer’s Alexia Nakad frames that as the real signal of brand seriousness, not the ad spend itself but the demand for the same measurement rigour applied to Meta, TikTok and Google. “As new channels like OpenAI emerge, marketers need to measure them the same way they measure established partners,” she said.
Enterprise thinking
But the enterprise data underneath all of this consumer-facing enthusiasm tells a more sobering story. Amir Grabic of Mahala.ai points to research that should give any CMO pause before greenlighting a big AI rollout: MIT’s Project NANDA found that 95% of enterprises see no measurable return on generative AI initiatives (MIT NANDA, 2025), and separately, only 7% of enterprises describe their data as fully ready for AI (Cloudera/Harvard Business Review, 2026). Grabic’s point isn’t that the technology fails, it’s that the plumbing underneath it usually does. “MIT traced the cause to integration and foundation issues, not model quality,” he said, citing the insurance industry’s habit of quietly adjusting premium data for negotiated client discounts as exactly the kind of unclean, siloed data that breaks a pilot the moment it tries to reach production scale.
So: are brands using AI? Overwhelmingly yes, but selectively, and mostly where it’s least visible to the end consumer. Behind-the-scenes production efficiency, measurement infrastructure, and adaptation at scale are where the real adoption is happening. The AI-first, fully-synthetic hero content Fiona Menzies experimented with remains the exception that proves the point: brands will use AI everywhere except the one place it’s supposed to save the most money and deliver the least return on the effort required to make it convincing.
