The Integrated Agency Failed Because the Money Was Never Integrated

Around 2015, during a WPP management development programme, a group of colleagues from across EMEA worked on a simple question: what should the agency of the future look like?

The answer was not complicated.

One client problem. One entry point. One brief. One team with access to the right disciplines.

That idea stayed with me because it was obvious. Clients did not want to manage creative, media, PR, digital, social, content, production and data as separate conversations. They wanted one clear way into the agency, and one connected answer coming back.

I also remember what happened after that.

We went back to our offices, our clients, our teams, our revenue targets and our P&Ls. I went back to a business I had to protect, revenue I had to defend, and work I wanted to keep inside my own walls.

That is why the integration conversation keeps coming back.

For years, agencies promised integration. The intention was usually right. The client need was real. The problem was that the model underneath the promise did not change enough.

Media had become more complex and commercially powerful. PR had built its own space. Digital arrived with different skills and economics. Social, influence, content, performance, CRM, data, commerce and production followed. Some of that specialization was necessary. Expertise deepened and clients needed sharper skills in areas that had become too complex to sit loosely inside one generalist structure.

But the split was never only about expertise.

It was also about revenue, margin, ownership and internal politics. Capabilities became businesses. Businesses needed leaders, targets, teams and commercial protection.

The result was a model that often made sense inside the agency, but made less sense to the client.

A client usually starts with a business problem. Growth is slowing. A brand is losing relevance. A launch is coming. Reputation is exposed. A category is changing. The board wants results.

The agency system often translates that one problem into several internal conversations. Creative, media, PR, digital, social and production may each have a valid view. The issue is that the client is often left to connect those views into one answer.

Clients asked me why one brief needed so many agency stakeholders.

More meetings. More alignment. More repetition. More effort to get to one answer.

They were paying senior fees and doing the agency's internal management themselves.

Clients often received coordination rather than integration.

People could sit in the same meeting, align on a presentation and appear as one team. Behind that, the machinery was still fragmented.

Coordination depends on effort. Integration depends on design.

The uncomfortable point is commercial.

True integration is difficult when every discipline is priced separately, measured separately and protected separately. If media has its own number, creative has its own number, PR has its own number and digital has its own number, then the client problem is already being filtered through the agency's commercial structure.

The system behaves the way it was designed to behave.

Clients also played a role in this.

Many clients separated creative from media. Procurement pushed pricing into line items. Different internal departments owned different budgets. CRM sat with technology. Digital sat with e-commerce. Media was audited separately. The CMO did not always control the full system shaping the brand.

Fragmentation was not only sold by agencies. It was also bought by clients.

Agencies fragmented to protect and grow revenue. Clients fragmented to control cost.

Both sides helped create the model they later found frustrating.

This is why AI matters, but not in the way it is usually discussed.

AI will make many things faster, easier and cheaper: analysis, production, adaptation, reporting, content creation, workflow and automation. But AI does not fix a fragmented operating model.

A fragmented agency with AI becomes a faster fragmented agency.

The real question is whether the agency is organized around the client's business problem, or around its own disciplines.

Publicis is the case everyone reaches for.

Its story is bigger than AI investment. The more important story is the longer direction of travel: Power of One, Sapient, Epsilon, Marcel and now CoreAI. Publicis moved earlier toward a more connected client model, supported by data, technology and simpler access to capabilities.

It also did not erase all its brands or disciplines. Leo Burnett, Saatchi & Saatchi, Zenith and Starcom still exist. The lesson is not that integration means removing every specialist identity.

The better lesson is that integration needs a stronger operating and commercial logic behind it.

The market has rewarded that direction. WPP is the useful counterpoint. It made major structural moves toward integration, merging agency brands and simplifying parts of the network, and the results were very different. Structure alone was not the answer. The commercial model was.

The new agency model cannot only be one entry point and one team on a slide.

It means one commercial logic around the client problem.

It means one P&L around the client relationship, not several discipline level P&Ls competing behind the scenes.

It means incentives linked to client growth, retention and value, not internal share of revenue.

It means specialists are rewarded for contributing the right answer, even when the right answer is smaller, different, or outside their own discipline.

The real test is simple.

Can the agency recommend less activity than expected, or a different solution than the one it usually sells, and still be properly rewarded for the judgment?

Almost none can.

And that is why the integrated agency has remained easier to promise than to build.

About the author

Habib Rihana

Habib Rihana is the founder of Short Pencils, an independent marcom advisory helping leadership teams, CMOs, founders, boards, and agencies make stronger marketing and communications decisions across the Levant and GCC.