RANCHO PALOS VERDES, CALIF. – The math sounds terrible at first: 90% of what an AI copywriting agent produces is garbage. But that still leaves 10 usable ideas generated in 60 seconds – a feat no human could match.
That calculation is reshaping how creative teams approach campaign ideation, turning artificial intelligence into a high-volume brainstorming partner rather than a replacement for human talent.
“Even if 90% of those lines of copy are garbage, those 10 might have some insights that a creative person can ingest and turn into something really impactful,” said Mel Stern, EVP, performance content, Omnicom Media, in this video interview with Beet.TV. “We’ve actually seen this happen on campaigns that we’ve worked on where a human and an AI bot are working together to drive creative ideation.”
Data-first creative development
Omnicom Media Group has been actively expanding its AI capabilities, including a January 2025 partnership with Google to develop AI-driven advertising optimization tools. The collaboration introduced planning solutions powered by Google’s Gemini AI.
For Stern’s team, the creative process begins with data rather than concepts. Audience insights, media performance data, and contextual signals inform the media plan first, with creative development following to bring that strategy to life.
“What we do is, being part of a media agency, is first and foremost, lean into audience data, media data, contextual data, any data we can get our hands on to inform our media plan,” Stern said, “and then say, now how can we make that plan come to life with creative?”
L-shaped overlays demand restraint
The proliferation of L-shaped ad overlays across connected TV is growing – from Amazon’s Thursday Night Football broadcasts to standard CTV advertising.
But Stern argues that emotional, anthemic campaigns from brands like Nike should avoid overlay distractions entirely.
The calculation changes for performance-driven efforts, where overlays can deliver personalization at scale by addressing different audience segments with tailored messaging.
“If you’re reaching, say, a sneakerhead audience versus a sports bettor audience, maybe you can use that overlay to speak to different elements of your brand, of your proposition, of your reason to believe,” Stern said.
But clutter remains the enemy: “You have to keep in mind there’s a base video running, there might be supers in that, there might be a VO, and you don’t want your overlay to distract from the main point.”
Personalization without the complexity tax
Dynamic ad servers, thousands of creative variations, and complex API integrations represent the heavy-machinery approach to personalized advertising. But Stern thinks brands dipping their toes into personalization waters can achieve meaningful results through simpler methods.
He recommends starting with basic hypothesis testing: compare creative that leans into audience behavior or contextual environment against generic alternatives. This would sidestep the cost burden of sophisticated dynamic serving while establishing proof of concept.
“Once we see it works, then maybe think about how do we evolve, how do we drive more complexity,” Stern said. “But not just because we can, but because we think we can actually have meaningful creative that will engage audiences.” The complexity, in other words, should follow the evidence rather than precede it.
