MediaSense is the firm in the room when the biggest advertisers on earth decide which agency gets their money. It is running Coca-Cola's global media review right now, which means somewhere in London a very large number of very expensive people are rehearsing a very large number of slides for the benefit of, among others, Ryan Kangisser, MediaSense's Chief Strategy Officer.

Ryan has watched more agency pitches than most agency CEOs have delivered. And his advice to every marketer walking into one is so simple it should be embarrassing for the rest of us: ask what they can't do, not what they can.

Think about how radical that is. The entire agency pitch economy is built on the opposite premise. The credentials deck is a monument to can. The capabilities slide is a wall of logos arranged like a family tree of royalty. The demo is a dashboard that glows like a slot machine and will be opened exactly twice after the contract is signed, both times by an intern looking for the export button. Nobody, in the history of Madison Avenue, has ever built a slide titled "Things We Are Bad At." Ryan thinks that missing slide is the only one that matters.

He would know. He sits on the other side of the table. And unlike most people who say the agency model is broken on LinkedIn, he has the survey.

Here's the number everyone in this business has quoted at a conference: in research MediaSense ran with the World Federation of Advertisers, only 11% of multinational marketers said their current media agency model was fit for the future.

Eleven percent. Congress polls better. Airline food polls better. Your cable company's hold music polls better.

I came in hot with that number and called it a complete vote of no confidence. Ryan, being British and therefore congenitally polite, corrected me on my own show, which I respect enormously. The 11% was about "fit for future purpose." Brands mostly accept the model they have today. They just don't believe it survives what's coming.

That sounds like a softer finding. It is not. A failing agency you can fire. A failing model you're stuck with. What the survey actually describes is a client base that has quietly decided the whole arrangement has an expiration date, and is now waiting to see who blinks.

The reason, in Ryan's telling, is that the clients changed and the agencies didn't keep up. After COVID, brands rebuilt themselves around direct-to-consumer, retail media, influencer, and now generative search, the whole alphabet soup of things that did not exist when most agency contracts were written. The client that emerged from that wants to "own more, do more, control more, be more independent," and wants an agency that mirrors all of that.

Read those verbs again if you run an agency. Own, do, control. The agency is not the subject of a single one. In the new arrangement the agency gets to be "additive," which is consultant English for "allowed to stay if it behaves."

The other famous number: three in four advertisers expect to change how they pay their agencies within three years. Great. To what?

Because "we want something different" and "we know what we want" are two different sentences, and this industry has been saying the first one since the Bush administration. The second one, never.

What everyone agrees on is the thing they're fleeing: time and materials, the billable hour, the ancient ritual of paying for how long something took. AI has made that model absurd on its face. The plan that took forty hours takes four now. The insight deck writes itself. The work is "not as dependent on lots of people and lots of systems anymore," which is a very gentle way of saying the client has read the same press releases about the AI platform that the agency put out, and has done the math on the invoice.

So what replaces the hour? Here's where Ryan gets genuinely useful, because he refuses to sell the fashionable answer. Everyone on the conference circuit loves outcome-based pay. It sounds righteous. Pay the agency for results! Who could object?

Ryan objects, and he's right to. His model is a blend: fixed fees to lock in strategic talent, some flexible commission where paid media still drives scale, and a growing slice of pay tied to output and outcome. Not a revolution. A mortgage refinance.

Pure outcome pay collapses the second it meets reality. If the agency handles media but not creative, social but not influencer, or the measurement is broken, or the product flops, or a war breaks out, the agency ends up docked for things it never touched. Do not fire the media buyer because the product team shipped a bad flavor.

And there's the data problem, which is the problem underneath every problem in this industry. Clients are not handing agencies the keys to the first-party kingdom, and they're not about to "overpay, in their eyes, for outcomes that may not have been generated entirely by their agency in the moment." Brands get built over decades, across all four P's, not by a single influencer post that happened to run the week sales jumped.

So: lots of appetite, very little action, "because it's actually really hard." Marketing departments spent generations getting comfortable with suppliers they could pay predictably and compare cleanly, especially during a pitch. Outcome pay breaks comparability. Breaking comparability breaks procurement. Breaking procurement is how people lose their jobs.

And then the thing that, from a pitch consultant, I did not expect: "Incentives drive behaviors." Get them wrong and the system loses "the level of provocation and agitation that marketers need."

Provocation and agitation. From the man running Coke's review. I may need to lie down.

Ryan wouldn't touch Coca-Cola specifics, and I'd have thought less of him if he had. What he would talk about is what a world-class advertiser wants in 2026 that it didn't want in 2020. The list starts with the thing that quietly went missing: strategy, which in his view has become "massively diluted over the last six years."

Pause on that. The pitch consultant to the world's biggest brands is saying the holding companies hollowed out the single thing clients actually pay a premium for. They spent the money on platforms with names that sound like Greek goddesses or cholesterol medication, and somewhere along the way the planners who could tell a CMO what not to do got reorganized into a shared service center.

What brands want instead is an agency that does what they can't, which Ryan summarized with the best line of the episode: why get a dog and bark yourself? They also don't want dependency. They want infrastructure that's interoperable and flexible, because nobody knows where this market goes next, and there are new releases every week.

Then came the sentence that should be printed on every holding company AI launch: "this is not a technology issue, it's an adoption one."

Buying the platform is easy. Every holding company has one. The hard part is getting tens of thousands of employees, spread across a hundred markets, carrying the scar tissue of a decade of mergers, to actually use the thing. What brands want is a partner that has the infrastructure and is "transparent with their own limitations."

Transparent with their own limitations. Ask what they can't do. It keeps coming back to the same slide nobody makes.