A thousand ads a week is easy now. Knowing which one deserves to run is the job.
AI can now make a thousand ads a week. The real work is choosing the one worth running.
Last week Runway made its new ads tool available to every paying customer. Connect your Meta, Google and TikTok accounts and it studies what's already working, then makes new versions in your brand colors and fonts. A week earlier, Higgsfield said its tool can produce up to 100 ad versions per product in minutes. TikTok and Meta announced their own versions of the same idea in October.
The number that stopped me came from Runway's own marketing team. They say they went from under 100 ads a week to roughly 1,000 with the same team, and cut their cost per subscriber by 41%. That's a vendor talking about its own product, so I hold it loosely. Still, it shows where things are heading. Making ads is turning into a cheap, endless supply.
My view is simple. When making things gets cheap, the value moves to choosing. Brands and agencies that treat this as a volume game will drown in it. The ones that win will treat it as a judgment game.
What the numbers are already saying
Look at WPP's latest results. In the third quarter, its production business grew 2.5% and its media business grew a little. Its creative business fell 5.1%, and Asia Pacific was its weakest region, down 4.3%. I don't read that as proof that creativity is dying. I read it as clients deciding they won't keep paying for creative work the old way, by the hour and by the piece, when a machine can make the pieces.
Look at who's buying whom, too. Lippincott just closed its purchase of Athletics, a studio that builds its own AI software for brand tools. The money is going toward people who can design the whole system: the idea, the rules for what fits the brand, and the tools that carry it out.
Closer to home, I liked a recent e27 piece about two Singapore agencies. One cut pitch prep from a week to two days. The other cut a brand analysis from 10 hours to one. Neither started by asking where to use AI. They found a slow, known step and fixed that. The same piece includes a story I can't stop thinking about: a weekly report meant to drop to one hour ended up needing four extra hours of fixing mistakes. Speed with no judgment behind it just moves the mess faster.
The honest case against me
Here's the strongest argument on the other side. For a lot of performance advertising, judgment may matter less than I think. If you can test 1,000 versions, the market picks the winner, and nobody needs a smart person in a room. Platforms already favor whatever gets clicks. Many small businesses will reasonably skip agencies altogether and let the tool run.
I think that's partly right. For simple, short term sales ads, more testing really will beat more debate, and we should stop pretending otherwise.
But testing only picks the best of what you gave it. It can tell you which of 1,000 ads got the most clicks this week. It can't tell you whether any of them made people trust you more, or whether they made your brand look like everyone else using the same tool. And people notice. Integral Ad Science found that openness to AI shopping tools among Singaporeans fell to 67% from 74% last year, even as most plan to keep spending. People are still buying. They're just getting choosier about what they let shape their choices.
What this means for brands and agencies
For brands, write down what your brand would never say or do, and make that the rule for every tool you use. A brand kit with logos and fonts isn't enough. The machine can copy your colors, but it can't know what you stand for unless you tell it.
Measure the right things. Cost per click will look great at first. Watch whether people still know who you are and why you're different six months from now.
For agencies, stop selling hours spent making things. Sell the idea, the rules and the final yes or no. Every one of these tools still needs a person to approve the ad before it goes live, and that approval is where our value now sits. Let's be honest that it's a different job from the one many of us were trained for.
And start small, like those two Singapore agencies did. Pick one slow step, fix it, and measure what it actually saved, including the time spent cleaning up after it.
There will be a thousand ads a week whether we like it or not. Our job is to make sure the one that runs is worth someone's attention.
Related reading and work
Cliff Lin
Writes The Monthly Ramble on AI, creativity, brand and social.
