Is AI killing creative tech?

The last time I wrote about creative tech in advertising (Can creative tech break out?) I had a pretty skeptical view, writing “Creative tech has been around for decades with very few winners.” That article was prompted (pun intended) by the rise in AI driven creative ad solutions, but really didn’t address the specifics of the issue of how LLMs and agentic solutions may or may not work.

That article was published in March, just four months ago, and I think my assumptions have changed quite a bit since then.

The unspoken assumption I had when I wrote that article was that ad systems (DSPs, ad servers, etc) were all going to incorporate AI to help create creatives for their customers, and that there would thus be little room to establish competitive advantage. I based this on the demos I was seeing which pretty much all went like this:

Step 1: Give us your website URL or upload some assets

Step 2: Choose a type of ad to create (don’t call it a template!)

Step 3: Here’s a not-that-bad ad, that you could use right now! Voila.

This workflow seemed good and right and useful, and largely still is. But then I saw this video created using Google’s Veo 3, and my mind melted a little:

Or this ad for betting site Kalshi:

The ability of talented creative individuals to build this level of quality 30 second spots using AI was not on my 2025 bingo card. It came a lot faster than I expected and while we’re probably a long way from every creative shop being able to do this level of work at scale, is there really any doubt that’s where we’re headed?

We have two extremely disparate data points and we’re trying to create a line. On the one side we have templated workflow that creates usable ads for unsophisticated buyers that otherwise wouldn’t have any ads at all. On the other hand we can create high quality ads for a fraction of the production cost when put into the hands of talented creators. How do we think these trends collide, and where will advertisers need help?

Benedict Evans in his newsletter often says something akin to “every SaaS company is just productizing an Excel spreadsheet.” A good point, but also you could probably make a different point circa 1990 that “every Excel spreadsheet is just productizing some paper calculations.” To extend the metaphor to AI, we’re currently in the paper-going-into-Excel era, rather than the productize-Excel-into-SaaS era. AI creative tech is taking activities that were done in Adobe Photoshop, Avid, Canva, studios, photoshoots, and really everywhere else and potentially replacing it with a blinking cursor in a prompt box. While Excel is one of the most widely used business tools, it has not historically been a good business model to sell plugins or templates for the spreadsheet — users tend to just figure it out themselves. Will AI also be a horizontal platform like Excel, or will companies emerge that solve the problem with more rigor?

Like Excel, there are wide variations in the skills being utilized within AI ad creation at the moment. The skills needed to create the Gorilla-Goes-to-the-Dentist video are substantial and creative people with experience in film making and design have a huge advantage and a critical role to play in utilizing this kind of technology. But as the models get better, you would have to expect the learning curve to decrease and the power to create to become democratized. How far away are we from a simple prompt like “here’s my brand, I want customers, give me five funny commercials based in Tucson, Arizona” and getting good results?

My bottom line is that AI creative development is moving ahead so quickly that it gives me even more doubt about whether there’s a viable market to sell tools for these use cases. Instead, I think AI ad creation will become as routine as using Excel in creative contexts and helpers will be embedded into lots of applications in the same ways you can routinely upload a CSV file.