Ad agencies are pouring hundreds of millions of dollars into a technology that threatens to reduce the amount they get paid.

But agencies have an answer to the conundrum. Campaign Red investigates. 

Underneath the headlines about major AI commitments is a complicated reorganization of talent and commercial models potentially representing the biggest shift to the agency landscape in its history, Campaign Red’s investigation into the impact of AI on agencies has found.

“We are in the middle of a transformation in how marketing services are being bought and sold,” says Stephan Pretorius, chief technology officer of WPP.

The fundamental problem, as Shufen Goh, Asia-Pacific president of MediaSense and co-founder of marketing consultancy R3, puts it, is that “agencies aren’t getting paid more for the use of artificial intelligence.”

On the contrary, marketers are now expecting to see the cost savings of automating many data and creative tasks reflected in agency fees.

For an industry that still charges by hours as standard practice, AI risks undercutting agencies’ ability to maintain revenue, requiring a rewrite of commercial models. Deliverable or outcome-based models are being explored, but they are complicated for clients to compare apples to apples. 

Agencies cannot sit on the sidelines, however, because there is widespread adoption of the technology among marketers. Nearly two-thirds of brand marketers (63%) are already using the technology, according to a recent report from the World Federation of Advertisers. Only 9% of brands last year said they have no current plans to use generative AI for their marketing. 

The research also found that 55% of companies surveyed were planning to renew media and creative contracts with partners to introduce AI-related clauses, and 48% are planning to introduce rules for how partners can use gen AI on their work. 

With no choice but to invest in AI or seem out of touch, agencies are balancing the commercial model risks by mostly reallocating existing funds into tech development rather than getting themselves into heavy debt.

With big shifts come winners and losers. Campaign Red explores…

Some of the world’s biggest global agency groups pledged in 2024 to spend more than a billion dollars on AI technology over the next few years, with Publicis, WPP and Havas all earmarking more than $300m each towards developing internal AI platforms.

While these substantial commitments seemed to spring out of nowhere, they aren’t entirely new funds, nor are they going entirely towards AI.

“It’s not like there was nothing and then suddenly we just whacked a whole lot of money onto AI software,” says WPP’s Pretorius.

WPP had a sizable fund for technology product development for “many years” prior to announcing its £250m annual investment in proprietary technology to support its AI strategy in early 2024, Pretorius says. The British holding company did not break out the size of its tech investment in prior earnings reports, but Pretorius says annual spend has “increased quite rapidly” since generative AI technology has taken the industry by storm. 

The agency network upped its spend to £300m for 2025, which now goes towards the ongoing development of its AI-powered operating system, WPP Open. This equates to approximately 9% of WPP’s net sales.

The largest chunk of WPP’s AI fund, according to Pretorius, goes towards employing the 3,500 people who are responsible for building software for WPP Open.

The next highest cost is cloud computing, with AI requiring significant computing power and vast, organized datasets to be effective.

Spend on enterprise licenses and AI APIs (application programming interfaces), such as those developed by Google and OpenAI, is a small slice of the overall investment – less than $10m a year, according to Pretorius.

That’s because these APIs are only useful to organizations when they are integrated into workflows and deployed organization-wide. “It’s an ingredient in a much larger software stack,” he says.

Havas also references talent, software partnerships with players like Adobe and Salesforce, as well as enterprise AI licenses, as key components of the agency network’s €400m, or approximately $428m, AI investment fund.

Arthur Fullerton, global chief technology officer at Havas CX, says the agency network is investing in reskilling its workforce to operationalize AI, a technology he describes as the “fuel” of its operating system, Converge.

“We’re not looking at AI as an opportunity to reduce our workforce. We want people to be skilled and be able to use these tools because we consider them required to do your job,” he says.

Kerry Howell, managing partner at VCCP’s digital agency Bernadette and AI agency Faith, estimates that 80% of the agency’s investment in AI is going towards people costs, with the remaining 20% on enterprise licenses.

Interpublic Group’s VP of emerging technology, Tom Sivo, says the network’s annual AI investment, which exceeds $100m, is “fairly evenly distributed” between two major areas: people costs, including hiring specialized talent, training existing staff and reorganizing teams; and licensing and platform costs, including cloud infrastructure. Smaller investment buckets go towards research and development (R&D), IT infrastructure and integration, he adds.

Publicis Groupe CEO Arthur Sadoun said last year that half of the company’s €100m ($163m) AI investment in 2024 would be focused on recruitment, upskilling and training, with the other half reserved for tech, licenses, software and cloud infrastructure.

‘Impossible to distinguish’ the true cost of AI

Other agencies don’t quantify these broader software stacks as AI spend, because they aren’t exclusively used to power AI. 

One global agency, which spoke to Campaign on background, said they spend around $10m per year on AI, consisting solely of the team allocated towards building AI software. They do not consider the cost of compute or API calls as AI investment, but the cost of doing business.

In reality, only a small portion of holding companies’ hundred-million-dollar investments in AI are truly incremental, since most of these funds go towards operating expenditures (OpEx) and capital expenditures (CapEx) that would have been spent regardless of the generative AI revolution.

IPG’s Sivo says the company’s AI investment “represents a combination of both net new funding and strategic reallocation of existing technology budgets,” adding that a “significant portion” comes from “newly approved” CapEx.

Part of the challenge in slicing out spend on novel AI technologies is how widespread the technology is being adopted across businesses, touching various budgets from innovation to R&D and IT. 

For example, several agencies cite Adobe and Google as their biggest AI partners, since these major technology platforms have baked generative AI tools and assistants into their suite of business services. But AI is not a separate line item in the multi-million-dollar contracts agencies sign with these companies.

“When you look at the investments that the holding companies and even mid-size and independent agencies are making in AI, a lot of the investment is embedded in existing technologies,” says Jay Wilson, VP, analyst at Gartner.

Wilson says AI represents “a set of technologies that are so much broader in scope than anything that has come before.” He adds: “It permeates everything from CDPs [customer data platforms] to content creation engines and everything else.”

For this reason, it is “impossible to distinguish [how much of WPP’s budget goes towards generative AI] because all of our tools and all of our workflows are being infused with AI from the start to the end,” says Pretorius.

Dimi Albers, chief executive of Amsterdam-headquartered marketing services company Dept, says the company no longer has a separate AI budget because “it became very difficult to literally track how much are you putting into it – since it’s basically become part of running the business.”

A difficult balancing act

Further complicating how to calculate the cost of AI for agencies is countering upfront investment in staff and tools against efficiency gains afforded by the technology. For example, handing off video production to AI means “no shoot, no CGI, very little post,” making it “way faster, way cheaper than human output,” says Wesley ter Haar, co-founder of S4 Capital firm Monks.

This will ultimately lead to a shrinking of the agency workforce, ter Haar expects, with less need for middle-layer roles like project managers, creatives and strategists. Instead, he thinks humans will live either at the most senior level of agencies – creative directors and account directors and heads of innovation – or “below the fold”, acting as assistants to AI, tuning models and implementing workflows.

In the short term, though, personnel costs at the big agencies have been going up. Salaries and related costs went up an average of 1% in 2024 compared to the prior year across five holding companies, according to their annual reports – whereas the number of employees at five of the big six holding companies fell overall, although figures varied significantly across the companies. Dentsu does not track salary costs or headcount at an equivocal level.

With AI opening up efficiencies in the creation and delivery of advertising, marketers are increasingly expecting their agency partners to pass along those cost savings in the form of reduced fees, according to many agency heads who spoke to Campaign Red.

Standard agency contracts are billed by hours on a sliding scale of rates depending on the seniority of the contributor. 

AI undercuts this model significantly, reducing both the hours a contributor needs to spend on client work and the number of contributors needed to service each client.

For agencies, it is a double-edged sword: they must show clients they are investing in AI in order to appear innovative, but expect their margins to be squeezed as a result.

Some agency leaders tell Campaign they are able to maintain the value of contracts by offering a larger volume of work  – thousands of asset iterations across a full spectrum of media placements – at a faster rate than before.

John Elder, CEO and co-founder of Supernatural AI Group, parent of AI-enabled creative agency Supergood, says the agency’s rate card can be similar to larger, more established competitors because clients are willing to pay a premium on “getting to KPIs faster.” Using techniques such as testing concepts with synthetic audiences in place of traditional quantitative and qualitative testing, he claims they can deliver a campaign in half the time of a large holding company agency.

“We don’t hear that we’re a lot cheaper, but we do hear, ‘Wow, you guys are so much faster and do so much more,’” he says.

That would have satisfied marketers earlier in the gen AI revolution, says Gartner’s Wilson. Now, with swirling economic pressures causing a shrinking of marketing budgets, and agency fees taking the brunt of the hit, marketers “expect to see the cost savings” reflected in agency rates, he says.

“Twelve months ago, CMOs were saying ‘we recognize that gen AI is going to enable you as an agency to deliver some cost savings to us, but we also recognize that you’re probably not there yet.’ So rather than demanding cuts to scope and reduction in hourly fees, CMOs expected to see the benefit of that automation coming through in terms of more creative concepts, more executions, for the same price,” Wilson says.

“Now, because of the budget pressures and the economic turmoil, it has very quickly shifted to ‘you need to shave down your staff plans and your related fees,’” he adds.

A rapidly changing business model

While account reviews that are handled by procurement still favor time-based models that are easier to evaluate on a like-for-like basis, some agencies and their clients have already shifted towards pricing based on outcomes and outputs, prompted by a desire for greater transparency and accountability.

“We have been helping a lot of marketers explore the impact of AI on compensation models, including moving towards a deliverables-based model,” says R3’s Goh.

“Outcome-based models are what everybody’s looking for, and I think this is a place where agencies actually have an opportunity to make an aggressive gain,” says Dept’s Albers.

These pricing models better enable agencies to push additional services to clients, balancing out the cost savings of AI.

Several agencies actually claim AI is enabling them to grow their revenue with clients, even with billings per work campaign going down, because clients are more likely to expand scopes when they see the benefit of automation.

“It’s not a question of our fees decreasing, it’s a question of charging differently – charging for outcomes, charging for outputs, and charging for licensing technology,” says WPP’s Pretorius.

The ultimate goal for agencies investing in internal AI systems is to transition into software providers, but licensing revenue is a “very small” slice of the pie right now, says Pretorius. 

“Small but growing,” adds MediaMonks’ ter Haar.

Outcome or deliverable-based pricing models lend themselves better to project-based scopes, a work arrangement that is taking over retainers as more work becomes AI-enabled. GroupM predicted that almost 70% of advertising was “AI-enabled” in 2024, a proportion that is expected to surpass 94% by 2029.

One major use case for AI is “tackling the creative grunt works”, such as resizing layouts or adapting designs, freeing up creatives to “focus on the big, impactful ideas,” says Wimala Djafar, director of H+, an innovation business unit within Hakuhodo International Indonesia.

Where agencies used to charge a set amount, say $50 per each image adaptation required for a campaign, it would be “completely insane” to use that pricing structure for an AI tool that can convert one master asset to 100,000 variants in seconds, says WPP’s Pretorius. This is an example of how AI is “completely rewriting the unit economics of many areas of marketing services,” he says.

The collapsing staff pyramid

AI is also collapsing the staff pyramid on client teams, requiring fewer workers overall but a higher concentration of senior staff, shaving out the middle management.

“How we’re discussing this with clients is you have fewer people, but very senior people, says ter Haar. “You get an amazing creative director on this specific piece of your business full-time, and a great strategist and client person and innovator. So instead of the 40 people that were normally on your business, that’s now 10.”

Senior talent have high salaries, which helps to justify keeping fees broadly the same or slightly lower, despite the cost savings AI enables.

This less but higher salary model may face pushback if marketing budgets continue to drop, says Gartner’s Wilson.

“We’re hearing that CFOs are increasingly getting involved in agency budget negotiations, and that when a CFO sees a rate card with a $500 an hour chief creative officer, and they combine that with a relatively nascent knowledge about what gen AI can do, they are questioning why they are paying for this person,” he says.

Agencies are also arguing that strong creative and strategic human skills have never been more important to guide AI into unexpected places and protect brands from derivative outputs.

“The opportunity cost is something I’m always talking about – the ability to do something you couldn’t do before,” says VCCP’s Howell. “In the same way that really great work doesn’t come for free because it’s the humans driving the process, it’s our humans that are really brilliant AI practitioners driving the creative process and using these tools, and that’s what makes AI brilliant. Of course, anybody can use it, and then it can be free, but you won’t get the same output.”

VCCP has developed an attribution model for AI when it includes the expertise of its AI agency, Faith, in a pitch, calculated as a contribution of time, expertise, and opportunity cost. Some clients push back against it because “they're not quite ready for the investment yet…but we have to put a number on it,” says Howell.

It boils down to how companies measure the return on investment of AI. Where marketers are primarily interested in cost savings, agencies are incentivized to show clients how deeper insights and speed to market are worth paying a premium for.

Some agencies say they can directly attribute about 10% of their annual revenue to AI right now, showing the speed at which the technology is taking over advertising. 

The challenge for agencies is to balance the power of automation with the value of human oversight, and find the best way to charge clients for that offer. It is a critical challenge; after all, the industry doesn't want to talk itself out of a job.