The Agency Model Didn’t Die. We Killed It Ourselves. And Now We Need What It Knew.

We are in the process of nailing the coffin shut on the ad agency model at just the moment when we need to relearn how to do what it was best at.

That’s not a defense of the old agency model. Big agencies and their holding companies earned their obsolescence. They became bloated, couldn’t keep pace with their clients’ needs, and often seemed more invested in protecting their retainer than solving genuine marketing problems. When the media landscape fractured and the platforms democratized distribution, agencies responded by pumping out the same work for different screens. They brought the logic of the thirty-second spot to a world that had already moved on, and charged for it accordingly. Clients were right to leave.

But what replaced the agency model wasn’t better thinking. It was massive volumes of shallower thinking rendered through better dashboards. At first it worked. Now it doesn’t. And we seem desperate not to have noticed.

For the last fifteen years, business has been run on KPIs. Every department — every person — became responsible for living up to a number within the data-fueled noise, often justifying another round of funding (and always justifying whether an individual was contributing enough to the organization) while obfuscating the metrics that actually mattered. Companies spent vast sums reducing churn, maximizing cart size, expanding install bases, optimizing for ROAS, increasing LTV, and lowering CAC, and remained shocked when those aggregate successes didn’t increase revenue or profit.

Lo and behold, now we put AI-powered data scientists in charge of marketing analytics and without the mandate to justify ad efficiency to a CMO or the existential need to prove that Meta ads are working, they are telling us bluntly and without sentiment that all these vanity metrics do is distract you from the real levers of growth.

If we’re capable of listening to the tools we’ve so eagerly laid off our staff in favor of, we would hear that the conditions that made transformational marketing possible are coming back.

Built for the Battle of Wits

From the 1950s straight through to the new millennium, the rules were simple. Small businesses printed fliers. Medium-sized businesses bought radio time. Large companies hired an agency. The hierarchy made sense because media moved slowly, it was highly relationship driven, and the audience was captive. Communication was one to many, the channels were finite, and the brands that could afford to dominate them did.

Agencies battled for attention with the only ammunition that wasn’t a commodity: their creative. Clients paid for the privilege of working with the best they could afford, and saw their fortunes improve, or decamped across the street to another firm. The arrangement was costly, but it worked for everyone.

In the 2000s, the dot-coms bubbled and burst just quickly enough that everyone but the sites themselves survived it, and suddenly there was a sense that the rules of building a business had changed. Reach and scale were available to anyone with a computer and an idea. Those watching Pets.com implode weren’t put off by its failure, they just got hungry for whatever was going to happen next.

Facebook fed them.

Riding the Rollercoaster

Facebook did something that no media platform had ever done before: it made a tiny budget feel like a superpower. In 2013, a founder with $500 could bin the flier and reach exactly the right person at exactly the right moment, watch them click, and see them checkout in a single session. It required no agency, no media buyer, no upfront commitment, and no persuasive argument. Just a clean, closed loop from impression to purchase, with no wasted spend, that seemed to invalidate everything the old model was built on.

We got drunk on it.

And in our stupor we lit a century of consumer behavior research on fire because ecommerce was novel and exciting and the numbers were intoxicating. We knew that building trust took time. That purchase decisions are emotional before they are rational. That repetition and context and brand feeling matter as much or more than the discount in the ad and the hour of the day. We knew all of this, and we still threw it out because for a brief, strange window, it seemed like none of it mattered anymore. People were clicking. People were buying. The funnel worked. Small brands were growing and the VC money was flowing.

The window closed. It always closes. Consumer demands didn’t change — they rarely do — the novelty of ‘new’ just wore off. Ads got more expensive as every brand on earth piled into the same auction and the platforms themselves capitalized on the monopolies they had built. Organic reach evaporated. Conversion rates declined. CAC climbed. And rather than conclude that the window had closed and reckon honestly with what that meant, the industry responded by going deeper into the data, hunting for the signal that would restore the magic. More metrics. More attribution models. More channel-specific KPIs. More dashboards.

Everything changed except our insistence that the system remain tilted in our favor.

Better Dashboards, Weaker Brands

What replaced the agency model wasn’t a miscalculation, at least not at first. The brands that blew up on Facebook and Instagram in the 2010s didn’t need a traditional agency. They needed cheap content, fast iteration, and a platform that rewarded both. Performance marketing served a real need for an explosion of smaller, faster, more numerous businesses that could never have absorbed the overhead of a traditional AOR relationship. The agency model needed Coca-Colas and Microsoft, and the world was suddenly full of direct-to-consumer startups with $2 million in seed funding and a very good Instagram grid.

But something got lost in the urgent need for more data and greater efficiency. When every dollar had to be justified against a channel-specific KPI, nobody was accountable for the brand itself.

Not the logo, the idea. The feeling. The vibes.

The reason someone reaches for your product or goes to your website when your ad isn’t in their feed. Brand building has low ROI when you’re reviewing spreadsheets weekly, so it just doesn’t get funded, and therefore gets ignored. The modern marketer became a day trader, buying and selling and pivoting and measuring, and ultimately falling short of the investor who put their money on a smart bet and let it ride.

The Machines Are Telling Us What We Already Knew

Now as we perpetually look to cut costs, we’re handing our marketing programs over to AIs, and they’re immediately pointing out what everyone has been so reluctant to hear.

Invest in brand even when you can’t measure it day in and day out.
Focus on revenue, margin, and growth rate in the aggregate.

These models don’t have a fiefdom to protect. They don’t need TikTok to be working. They have no need to prove that their investment in audience growth justified its budget or that the attribution model their team spent a year building deserves another quarter of funding. They’re looking at the entire picture and telling us that the four dozen KPIs we’d carved our businesses into were nothing but beautifully organized chaos.

They’re saying that even with their vast computing power it is a fool’s errand to try and measure the impact of every individual ad. That what moves a business is awareness and excitement, and that most of what we’ve been optimizing for has had a tenuous relationship with either of those things at best.

They’re telling us in no uncertain terms that the party has been over for a while. We’ve just refused to go back home.

Simple Math for Massively Complex Times

These AI models won’t save us from ourselves. Ask them for an optimized Meta model and they will deliver the most optimal Meta model they can find, which is not the same thing as the most optimal marketing model. The machine answers the question you ask. Knowing the right questions is the hardest part.

When you do ask the right question, what these models are telling us is what we always knew and chose to forget. People are messy. They resist and sabotage any serious attempt to be tracked, measured, and predicted. Performance marketing does not work the way it used to. Brand building does work, but it works slowly and across channels in ways that don’t always fit cleanly together. Almost everything that promised a shortcut to growth was really just a shortcut to a number that looked good in a board meeting. And omni-channel, creative marketing over a longer time horizon delivers actual business results.

The computers are doing what computers do best: ingesting data, computing probabilities, surfacing what the numbers actually say.

It’s time to let people do what people do best. Tell stories. Build relationships. Make something worth remembering.

The real irony is that when you cut through the bloat and the lousy corporate stewardship, this is what ad agencies were best at. Big ideas. Brand building. Short-term velocity. Long-term impact. At exactly the moment that we are battling AI slop, deepfakes, and impotent content, we are writing the eulogies for the only ones who knew how to rise above them.

The good news is that while the institutions may have collapsed or pivoted into something unrecognizable, the people who made the work what it was are still here and they are hungry to get to work if we will only give them the chance.

Creativity inspires. Inspiration drives affinity. Affinity drives revenue.

Simple math for massively complex times.