Creative drives most of the result. The industry books it as non-working. The AI tools are commodities, the data is the moat.

A lot of it was opportunity. Collmer had been running a video game business he built with two other people, and as its advertising moved from television to digital channels like Facebook, he saw the web turning from static text and images to video, with enormous friction ahead. If everyone would struggle to make the content the medium needed, that sounded like a business opportunity. Disillusioned with entertaining 14-year-olds, he wanted meaningful work, and believed creativity was one area that would be protected as technology pressured labour. So he set out to build a marketplace matching creative talent with businesses that needed it.

It is funny now, he says, because AI's ground zero turned out to be creativity. But at the time he thought that was the protected area. That intersection of technology risk and art risk has been his whole career, and even as Vidmob has matured, the question is the same: how do we use technology to enhance creativity and make it a force for business growth?

It used to be very difficult to measure. Understanding everything happening creatively inside a piece of content was not possible even five or six years ago, so brands classified things at a high level: a comic approach, a promotional benefit. They measured whether an ad worked, but not why. That is the big change now. Technology can get at not two or three characteristics of a piece of content but tens of thousands.

It starts with a structured, repeatable taxonomy for a true data-driven view of creativity. A ten-second Instagram ad is a set of static frames, each with a logo, on-screen messaging, audio, pacing. Turn all of those decisions into data, compare them to equally rich performance data, and first human data science, then machine learning, then AI can parse what is really driving the behaviours you care about.

Outcomes have broadened from fast media metrics to sales and brand behaviours, and the fast metrics still have utility because of their speed. But there is a never-ending debate between brand and performance, and for a while performance has won it, driven by immediacy, because everyone has quarterly results to hit. If something is useful but won't have material impact until later, it is hard to invest in properly.

As we create more and more content, Collmer argues, more and more of it becomes unmemorable, and performance noise becomes invisible unless a brand is built with purpose. MMA research, he says, shows today's brand investment lifts future performance by multiples two, three and four quarters out. Selling that to a CFO is where good CMOs are divided from bad ones.

The most exciting launch was Vidmob's MCP in June. Every client now has dev teams or individuals playing with AI, and giving them access to properly structured data makes what they build instantly better. Vidmob has moved away from getting everyone to experience everything inside a static UI, and towards helping them build a unique data asset that makes any AI-enabled tool better, across planning, briefing, asset creation, media decisioning and measurement.

As a product company the focus used to be on the experiences and the end product. Collmer has taken all of Vidmob and turned it inside out, under a holistic framework called Vidmob 360, the unifying name for its MCPs, APIs, recommendation engines and AI tooling. He has been focused on what Vidmob looks like when it is just something that makes your thing better, rather than the thing itself.

As amazing as these AI tools are, Collmer says, they are ultimately commodities. It won't make a difference whether you use Claude, Gemini or OpenAI, or one generative model or another. What makes a huge difference is knowing what to make and the input you feed in. The world is misfocused on chips and algorithms. He thinks data is where one business will beat another.

On Meta's Global Creative Council, Jimmy Stone has talked for years about compound branding. It has become easy for everyone to create, driven by generative AI and by creator content, which Collmer calls a human generative AI business. Together they cause a loss of control and even awareness of how a brand is represented. The question becomes how 10,000 or 50,000 pieces of content operate in concert around the purpose of a business, rather than a bunch of people in their bathroom saying put on this lipstick.

If there is no memorable takeaway, it is just noise. Collmer watched game publishers put hundreds of billions into app install campaigns, making anything and throwing it at the algorithm. He called them shit throwers. It drove enormous installs, but nothing else, and when the money stopped there was no brand left. Many of those businesses no longer exist. Unilever's extensions of its Real Beauty campaign, by contrast, show how to build unifying campaigns anyone can plug into.

For someone starting out on the creative side, Collmer says it is no different than before: you still have to be great, striking and differentiated. AI, like the printing press, let more people produce more, but it didn't make anyone a better writer. If you want to stand out, you have to use commonly available tools in uncommon ways. He has seen beautiful work made with AI, and he has seen people say make me an ad to do this and get what they deserve.

The second risk is sameism and atrophy: say write my email and a month in you don't remember how to write one. This is a period, he says, to learn how to maintain your skills and humanity without refusing to touch the tools. And the person receiving your communication has to feel it came from you. When he can tell a brief was written entirely by Claude, his willingness to read it goes to zero.

Collmer believes the broader your input, the more useful it is. Podcasts across business, history and psychology, consumed on long runs that give him alone time, feed the part of his job that is figuring out where the world is going and keeping Vidmob positioned better 12 months out than today. You get there with a broad focus rather than a narrow one, and the best ideas often come when he is not even thinking about the task.

He points to Coca-Cola as one of the most leaned-in, earliest adopters of creative data, impressive in how it looks at the whole-scale understanding of its brand embodiment, almost in a world of its own there.