S4 Capital's Monks is pushing subscriptions. WPP is pushing outcomes. Meanwhile, the real question goes unanswered: who's actually paying for all these AI tokens?

The billable hour is dying. Everyone agrees on that. What nobody can agree on is what should replace it.

Last week, two competing visions emerged from opposite ends of the holding company spectrum. S4 Capital's Monks announced it expects roughly a quarter of its revenue to come from subscription fees by year end. Meanwhile, WPP CEO Cindy Rose declared that outcome-based compensation—getting paid for actual business results—is "the beginning of a more widespread commercial model evolution."

Both are trying to solve the same problem: AI has broken the fundamental economics of agency work. When a thousand ad variants cost what five used to, charging by the unit makes no sense. But the proposed solutions reveal very different assumptions about where the real value lies.

The Subscription Play (And Its Problems)

Monks' pitch is deceptively simple. As AI moves from pilot to scale, agencies are racking up real expenses—inference costs, token budgets, licensing fees for tools like Runway and Claude. Clients won't accept sudden surcharges on an invoice. So wrap those volatile costs into a predictable monthly fee.

"The cost of inference becomes a line item in the P&L instead of just hidden in a bunch of little pilots," explained Monks co-founder Wesley ter Haar. The agency negotiates bulk token deals with AI providers, runs client work on that discounted infrastructure, and charges a subscription that covers costs plus margin.

Critics have been brutal. The loudest charge: this is just a retainer with better branding. And there's a more serious concern underneath—what some are calling "token arbitrage." When an agency manages AI behind a subscription, the temptation is obvious: run cheaper models, keep the margin, and the client never knows.

"It's the same dynamic that caused so much trouble with agency trading desks a decade ago," warned Robert Webster of TAU Marketing Solutions. Back then, agencies quietly profited from media buys clients assumed were made purely in their interests. Now the arbitrage is computational rather than media-based, but the information asymmetry is identical.

The Outcome-Based Promise

WPP's approach is more ambitious, and therefore riskier. Rose wants to tie agency fees directly to business results—sales, brand performance, measurable outcomes. The clearest proof point: Jaguar Land Rover, where WPP is negotiating a global creative and marketing partnership with fees linked to actual vehicle sales rather than hours worked.

"Those outcomes aren't 'do you like the agency you work with,'" said Johnny Hornby, CEO of WPP's specialist communications division. "Those outcomes are 'are we selling more product and will we get paid on being able to sell more product?'"

The infrastructure exists, sort of. WPP's data platform, built partly around InfoSum's federated technology, can connect first-party data, media signals, and sales outcomes without data leaving a client's environment. For Heineken, that meant linking shopper data with ITV viewing audiences and Tesco sales figures to measure real in-store uplift.

The appeal is obvious: clients get a partner whose incentives match theirs, and agencies get paid for value created rather than hours logged. But the gaps are also obvious: What happens when targets are missed? How are outcomes independently verified? WPP shared no specifics on what percentage of revenue currently sits under any outcome-linked structure.

The Uncomfortable Reality

Here's what neither model addresses: 85% of Monks' AI work still has humans driving it, from initiating tasks to checking outputs. That's not an always-on autonomous system. That's people doing their jobs with better tools. Charging a subscription premium—or staking your fees on outcomes—before the AI is genuinely doing the heavy lifting is getting ahead of what the technology can honestly justify.

Meanwhile, WPP's CFO Joanne Wilson admitted the company is running three commercial tracks simultaneously: outcome-based pricing, output-based pricing, and technology licensing. "With many of our clients, we're working to understand what works best."

Translation: nobody knows yet.

The real prize, if it exists, is further out. Build always-on infrastructure, absorb variable costs, use accumulated performance data to make renewal easier than re-pitching, and eventually prove a causal link between agency work and business results that justifies genuine performance fees.

Get it right, and the commercial benefits cut both ways. Get it wrong, and it's retainers with a better story and a new margin-hiding mechanism. The industry hasn't figured out which version it's going to be yet.

But at least they've stopped pretending the billable hour has a future.