ERCOLE▮ AUTOMATIONS ← All Thoughts
Performance Marketing · AI Systems

Is the Performance Marketing Agency Dead? Let's Discuss.

Justin Ercole● July 6, 2026● 7 min read

The performance marketing agency likely isn't going away anytime soon, but the model that built it, charging a premium on retainers to run what a single AI-equipped operator can now replicate, has a short shelf life.

I spent a lot of time digging into this because I needed to understand where I fit. I build AI automation systems for brands and content creators, but performance marketing was my background and where I came from before moving fully to building my own AI automation systems. And there is a lot of noise right now about agencies being dead, AI replacing media buyers, and the whole industry falling apart. I wanted to know what is actually happening versus what people are just saying.

The short version: agencies are not dying. The commodity middle tier is. And the value chain has moved in a way that most people have not caught up to yet.

What I found

The production cost floor already collapsed

A peer-reviewed study out of MIT found that AI assistance cuts professional writing task time by roughly 37% and improves output quality by around 18%. That is one data point, and it applies to writing specifically. But the directional trend holds across every creative discipline I looked at. What used to take a team and a real budget now takes a prompt, an afternoon, and a good workflow. The question of whether AI creative is "good enough" is largely settled. It is. For a lot of use cases, it is better and faster than the old process.

This matters for agencies because a huge portion of their retainer fees were justifying the overhead of creative production. That justification is getting harder to make.

The platforms absorbed what agencies used to charge for

Meta Advantage+ and Google Performance Max now handle audience targeting, bid strategy, and placement optimization automatically. These were the things that justified 20 to 30% management fees. The platform does them by default now, and it does them with data no third party can match.

Industry analyst data shows that global ad spend growth is outpacing holding company revenue growth. The delta is not flowing to agencies or to operators. It is flowing straight to the platforms. Google and Meta are capturing the margin that used to be distributed across the media buying layer. That is a structural shift, not a dip.

Agency revenue held. Agency profit did not.

Here is the thing that surprised me most. The US marketing agency market held at $182.5 billion in 2025, growing at about 5.5% annually. Agencies are not going out of business. But their margins are getting squeezed from two directions at once. Clients are pushing back on hourly rates because they know AI is making things faster. Junior roles in reporting, admin, and production are becoming redundant. A traditional agency runs on 15 to 20% net margins. That math does not work well when your biggest cost line is the labor AI is replacing and your clients are demanding to see the savings.

The Forrester 2026 agency predictions point to real structural consolidation. Havas acquiring dentsu's international operations. Omnicom and IPG mergers. 85% of US B2C marketing executives planning agency reviews in 2026. Revenue holds. Headcount and margin do not.

This exact thing has happened before

Desktop publishing hit in the late 1980s. Aldus PageMaker made it possible for designers to do what typesetters and paste-up artists had done for decades. The production specialists did not adapt fast enough. Within a decade, that entire layer of premium service shops was gone. The Harvard Business School documented the same thing happening with programmatic advertising in the early 2010s. Mid-size media buying agencies lost accounts to in-house trading desks and to holding companies that had built proprietary data layers. The Harvard case put it bluntly: it was "seen as unlikely this business will return to the media agencies."

Every one of these disruption cycles follows the same arc. A new tool collapses a production layer. The middle tier of generalist shops that competed on that production layer gets crushed. High-trust, high-accountability specialists survive. The people who survive are not the ones who mastered the tool fastest. They are the ones who used the tool's arrival to redefine what the work actually was.

The value chain moved. Most people have not caught up.

The question in performance marketing used to be: who can make the most variants, fastest? That had a high cost and a real moat. Now it has a cheap answer. The new question is: who owns the data, builds accurate measurement, interprets the signals correctly, and carries accountability for the outcome? That is a much harder thing to automate.

There is also the trust layer. A recent AdExchanger interview with Jonathan Roberts of People Inc. put it simply: people do not trust AI, they trust creators. The human editorial layer, the person or brand with a real point of view, is not a production cost. It is the thing audiences actually follow. That is a different kind of moat than creative output speed.

The thread that connects it

AI didn't just lower the production cost. It moved the entire value chain.

Every part of this story points at the same structural shift. The work used to be making the thing. Creative, media buying, reporting. All of it had a labor cost attached and a team to match. AI and the platforms have made the making cheap. The new work is knowing what to do next. Who owns the data. Who builds the measurement infrastructure. Who carries the risk when the numbers do not move. That is where the value is. And most agencies and most operators are still competing on the layer that just got commoditized.

What I'm still watching

The measurement and attribution problem is the piece I keep coming back to. iOS privacy changes removed a significant chunk of deterministic signal from Meta campaigns. Third-party cookies are going away. AI-generated creative is harder to test in clean holdout conditions. The entire conversation about "operators vs. agencies" assumes someone can accurately measure what is working. That assumption is shakier than it looks.

My bet is that whoever builds reliable, privacy-first, first-party data infrastructure for mid-market brands in the next 18 months will hold a more defensible position than anyone competing on creative output speed. The platforms can absorb targeting and creative. They cannot absorb the custom integration work that makes a specific brand's data stack trustworthy. That is where I am paying attention.

Sources

I am building at this intersection right now. Automation systems for brands that want performance without the retainer overhead. If that sounds like a conversation worth having, reach out.

Work with me →