Forrester Put Creative Services in the Same Risk Column as Custom Software

Forrester has a name for what marketing agencies have been doing with AI, and it coined the name back in June 2025. It calls it the AI cost center crisis.

The number under it is 75%. In research run with the 4A’s, Forrester found that “75% of marketing agencies bear the costs, with generative AI capabilities funded directly by the agency without passing the costs to clients.” Forrester called that “an astounding 83% increase from 2024.”

On August 19, Forrester published something that makes those two numbers read differently.

The scoreboard

The AI Disruption Model scores 17 technology and service categories across more than 200 markets. Analysts Craig Le Clair and Ted Schadler built it around nine drivers. Two of them matter here: how easily AI substitutes for the work, and how labor-intensive the market is. A third is the commercial model itself.

Creative services landed in the worst group. Forrester’s words:

“The most severe disruption stemming from AI hits labor-intensive markets, such as technology implementation, custom software development, creative services, and corporate training.”

That’s the category most marketing work gets sold under, sitting in the same sentence as custom software development, the market everyone already accepts is being rebuilt.

Forrester built the model to sort winners and losers across the whole technology and services economy, and it puts infrastructure, data and security in the growth column. Creative services scores badly on the inputs.

Why the two numbers collide

Here’s the part worth sitting with. Labor intensity is not something an agency chooses. The commercial model is.

Forrester’s June 2026 agency research found nine in 10 agencies using generative AI, and half using agentic AI. Agentic means the software carries a job through several steps by itself, instead of stopping for instructions between each one. The same research found 61% still classify AI as a “cost of business” with limited direct monetization. Only 31% plan to monetize it inside 24 months.

So a year after Forrester named the cost center crisis, most agencies are still the cost center.

The gains are not in dispute. The same Forrester work reports one agency increasing “speed to market by 80% or higher compared to its non-AI projects.” Others reduced “production costs between 40% and 50%.” Global brands applying AI end to end saw advertising performance rise “up to 70%.” None of it shows up in what gets invoiced.

Jay Pattisall, VP and Principal Analyst at Forrester, put the failure mode in one line:

“AI has fundamentally transformed marketing agencies, but the industry is at risk of mistaking efficiency for effectiveness.”

What it looks like from inside an agency

Robin Bonn, CEO of the agency consultancy Co:definery, described the sequence in a guest post for the 4A’s. The plan, as he tells it, was simple enough:

“Drop the headcount, decrease costs and maintain prices.”

It didn’t hold. Clients could do the same math. Bonn writes that “clients demanded those savings be passed on,” and lands on the condition that actually sets the price:

“Agencies are now seen as interchangeable other than by price.”

That sentence is the disruption model restated in plain English. Interchangeable-other-than-by-price is what a commodity is. Forrester’s nine drivers are a formal way of asking whether a market has become one.

The mechanism

An agency that cuts its own production cost by 40% and holds its rate has invented a margin, and margins invented that way have a short life. The client eventually learns what the work costs now. Procurement is very good at learning this.

At that point the agency has two options and only two. Charge for the output, which means selling something other than hours. Or hand the savings over and run a smaller shop on the same rate card. Most have picked the second by not deciding. That’s what “cost of business” means on a survey form.

This reaches the founder buying marketing services, not just the agency selling them. The second option has a visible symptom. Fewer senior people on the account. Faster turnaround, thinner thinking. The rate holds and the thing you were buying quietly changes.

The counter-argument

There’s a real one, and Forrester supplies it. Bonn’s own read is that pricing is the wrong place to start: he argues “agencies must prioritize proposition innovation before pricing-innovation,” and Pattisall’s recommendation runs the same direction, that agencies “must now reset expectations and invest in creativity, talent, and marketing performance.”

If that works, labor intensity stops being the whole story. A market scores badly on AI substitutability only for the work AI can substitute. An agency selling judgment rather than production gets scored on a different axis. The 31% planning to monetize inside two years are the ones testing whether that axis exists at scale.

It’s a live question. It’s also a small minority, and 24 months is a long time to hold a rate card that clients have already started pricing.

The verdict

Watch the invoice, not the announcement. Every agency has an AI story now, and the story costs nothing to tell.

One question sorts them: does AI appear anywhere on what they bill you? If it doesn’t, you’re buying from a business absorbing a cost it can’t absorb forever. The adjustment arrives as a rate conversation or as a quieter team. Ask which one is coming before your next renewal, because that ratio is already a published target at the holding companies.