For anyone who has worked in advertising long enough, today’s excitement around AI feels remarkably familiar.
Fifteen years ago, the mainstream programmatic buying boom transformed media planning almost overnight. Instead of calling individual publishers to place campaigns, marketers suddenly had access to nearly limitless inventory through a few clicks. They could reach more people, more efficiently and at a lower cost than ever before.
Of course the industry embraced it. But what we didn’t ask (or at least not quickly enough) was whether all of that inventory was worth buying.
It took years to untangle the consequences. Viewability standards had to be established because ads weren’t always being seen. Brand safety became an industry of its own because brands discovered their campaigns appearing next to decidedly brand-unsafe content. Invalid traffic and fraud consumed billions of advertising dollars before marketers built the measurement infrastructure to understand what was actually happening.
As an industry we were so excited by the potential of programmatic that we collectively ran out and bought the luxury car without test driving it. This time, let’s test drive before we buy, because scale alone won’t create better outcomes.
Every major platform is promising marketers the ability to produce more creative variations, more personalized assets and more content than any team could have created manually. In a market where budgets remain under pressure and expectations continue to rise, the promise to test more creative at scale is understandably attractive.
However, just because we can create exponentially more creative, doesn’t mean we should. Think about one of the most famous and oft-referenced advertising misfires of the last decade: Pepsi’s Kendall Jenner campaign. Exceptional production quality, enormous budgets, talented creative team, big name celeb…. What could go wrong? Yet the campaign failed because it fundamentally misunderstood its audience and the cultural moment. Panel tests measuring stated preferences weren’t enough to catch that, and that’s the real question worth asking: when you’re about to make a bet that size, with that much on the line, why wouldn’t you want every possible signal in your favor before you commit? The data to see around that corner exists now. In the end, the campaign flopped not because the production wasn’t great, but because a smaller, earlier investment in understanding the audience could have changed the outcome of a much bigger one.
So what are marketers risking? It’s not always about brand safety; sometimes it’s about brand dilution, a weakening of the consistency that brands spend decades building. Then of course there’s wasted spend and optimizing creative that doesn’t work. Every asset that misses the mark carries production costs and media investment.
To benefit from AI, marketers need to be asking smarter questions before their assets ever go live: Does this reflect our brand? Will it resonate with our audience? What evidence suggests it will perform? Are we scaling quality or simply scaling output? Most importantly, they should ask: “What evidence do we already have that this creative is likely to work?” before they invest in producing hundreds of variations.
Going back to the car analogy, no one would spend $100,000 on a new SUV without taking it for a test drive. Yet brands routinely commit millions of dollars to producing and distributing creative based on instinct, limited consumer panels or internal opinions, only learning what worked after the media budget has already been spent.
Fortunately, we can avoid these mistakes. Today’s marketers have access to historical performance signals that can reveal which creative elements consistently drive attention, engagement and business outcomes across thousands of campaigns. The smartest marketers are leveraging those insights to inform creative decisions before AI scales production, not after media dollars have already been spent.
Programmatic eventually matured because the industry invested in better standards, better measurement and greater accountability, but only after a reckoning forced its hand. The question for AI creative is the same one we should be asking now: after what cost, and how much lost time, will that reckoning come? Brands don’t have to wait to find out. There’s an opportunity to get ahead of that curve today, by leveraging creative intelligence to build the discipline in from the beginning rather than bolting it on after the damage is done.
