15 spots left to join us for the Digiday Publishing Summit
Join us Sept. 14-18 in Miami to connect with top publishing leaders
SECURE YOUR SEAT
AI Revenue Generation
S4 and Monks are trying to get ahead of rising AI token costs
Rising costs associated with AI tokens are forcing marketing services groups to confront hard questions about how much they should be spending on AI.
Token usage is “exploding” at Monks, according to co-founder and chief AI officer Wesley ter Haar. That’s to be expected, given the company’s client base of tech brands, production heritage and years-long investment in generative AI tools, but could become a
Parent company S4 Capital’s latest set of earnings, published last week, prompted a share price jump for the London firm; strict cost discipline helped broaden margins to 12.3% and double its first half operating profit to £35.2 million, up from £16.4 million during the same period last year. The numbers followed a positive tranche of Q1 figures.
Speaking to Digiday, S4 chair Sir Martin Sorrell called it a “huge improvement,” but acknowledged that rising token costs could demand a stricter spending policy. “If I was going to be critical, I think we’ve probably been ill-disciplined on that,” he said.
How the company manages that cost in the future is an open question. Goldman Sachs forecasts suggest token consumption could increase 24-fold between 2026 and 2030, a rise mostly driven by enterprise and business usage. Licensing fees covering the number of employees using an AI tool are also adding to outgoings.
Some marketing services companies, like indie agency PMG, have instituted caps on staff token usage. The Texas-based media shop established a $50 daily cap on employee token usage in May.
It’s ahead of the pack in taking action. “The other shoe has yet to drop when it comes to the cost of AI inside agencies,” said Forrester vp and principal analyst Jay Pattisall. He cited a recent Forrester survey that found 60% of agencies were prioritizing spending on third-party tools, usually paid forng cloud or compute costs
In any case, such a blanket policy may not suit agencies like Monks, which operate with a broader range of practice areas. “It’s probable that we’ve overused or overindulged. But given what we do, that may not necessarily be a terrible thing,” said Sorrell. “The reality is we’re going to have to spend more on increasing our capability,” he added.
Ter Haar suggested a “horses for courses” approach, in which caps might be set for teams using AI agents or tools for lower priority tasks, but lifted for other areas of staff expertise such as coding — a practice area that has seen token usage “explode” upwards as the use of agents has become more common.
“If you let your whole team tokenmaxx, then you’ll have a bunch of people sort of burning [dollars] figuring out what the weather is next week,” he quipped.
The company hopes softer limits will curb token usage without curbing staff effectiveness. Its flagship suite of AI tools Monks.flow, can shift between models per their suitability for a given task, a judgement that can incorporate token efficiency. And the company can restrict some staff to using set menus of AI agents within a prescribed workflow, keeping their token usage predictable.
Marketing agencies across media, creative and production are attempting to find the right balance between investment and outgoings, and between token usage transparency and complexity for clients. But these are especially difficult questions for S4, which achieved its broadened margin after years of staff cuts and cost discipline. The company, which employed over 8,200 people four years ago, now has a staff numbering 6,150.
Sorrell and ter Haar, and their peers at agency C-suites, will have to make a call either way before long, though. “The reason any of this is possible is because the equities markets and PE are subsidizing the costs of AI at the moment,” said Pattisall. “When investors wake up to the fallacy of this business model, and third-party providers switch from SaaS to a token or GPU-based model, agencies will find the real cost of AI to be significantly higher — and possibly more expensive to automate advertising then to hire humans to create it.
“This scenario could have a devastating effect on the marketing agency industry, which shed about 8% of staff in 2025 and the first half of 2026,” he concluded.
Per ter Haar, token costs are bundled with other costs for project clients and passed on directly, with no additional margin, for Monks.flow clients.
Outcome based pricing models may offer a means of scaling revenue as token costs rise. S4 has pushed ahead on both bespoke performance-linked fee models and a subscription arrangement.
The latter provides client with a set number of outputs and regular updates, for a fixed fee; Ter Haar said it’s on track to take 25% of its revenue from subscriptions by the end of the year. “Almost all new business is either subscription or [outcome-based],” he said.
Sorrell’s old vehicle WPP is also gradually embracing outcome based pricing, in part motivated by a need to account for altered cost considerations.
“In the past, our business and values really come almost entirely from people now, it’s people and tech costs,” CFO Joanne Wilson told Digiday upon the holding company’s earnings release. “So, as part of that, we’re evolving our commercial model so that we’re reflecting those inputs between people and tech.”
But ter Haar said that while CMOs often find performance or outcome based arrangements intuitive, it’s still a difficult sell to the corporate procurement teams involved in pitches or RFIs. Forrester’s research showed just 9% of agencies monetize generative AI, while 61% of agencies consider AI part of the cost of doing business
“I think clients aren’t good at buying variable commercial models yet,” he said. “If you’re going through a buying process… pricing at the rate card level is just the easiest apples to apples [comparison].”
More in Marketing
OpenAI is coming for SMB advertisers
The company is going after smaller advertisers. Its job listings prove it.
Zoom relies on creators to boost AI search results
As AI changes how brands are discovered, Zoom is leaning on creators and journalists to shape its reputation.
Your favorite beauty brands are now also entertainment brands
Many of today’s beauty brands are asking if entertainment and advertising can be one and the same.
