Future of Marketing Briefing: The case for and against an agency subscription remuneration model

Below is a case for and against subscriptions. 

The case for subscriptions 

It’s not about pricing, it’s about costs

The real problem subscriptions solve isn’t pricing it’s cost absorption. 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 — that clients won’t accept as sudden surcharges on an invoice. As Monks’ co-founder Wesley ter Haar put it: “The cost of inference becomes a line item in the P&L instead of just hidden in a bunch of little pilots.” Subscriptions are, therefore, his agency’s way of how to stop that from becoming the client’s problem. 

The infrastructure play underneath

But there’s more to the model than billing convenience. Monks negotiates volume discounts on tokens throughout enterprise deals with AI providers, then runs client work on that discounted infrastructure. The subscription fee is set to cover those costs plus margin. In other words, it’s the client-facing wrapper on top of whole infrastructure strategy. Bulk token purchasing and subscription pricing only make sense together: one creates the cost structure, the other creates the revenue structure to sit on top of it.

The work itself has changed

Then there’s the nature of the work itself. Agencies are increasingly building always-on systems — brand knowledge bases, trend monitors and tools that flag when a competitor does something worth reacting to. That work never really stops, which means it doesn’t fit neatly into a project invoice. As Robert Webster, founder of AI marketing consultancy TAU Marketing Solutions, put it: “What’s being built is infrastructure, not a campaign. And infrastructure can’t be billed the same way as a TV spit.”

This matters more given the wider shift in how advertising costs are structured. Media buying and creative production are both moving toward consumption pricing — pay per token, per output, or total number of impressions served for instance. Agencies caught in the middle risk getting squeezed unless they can wrap those volatile costs into something cleaner. Subscriptions are essentially that wrapper.

The longer game

There’s also a longer game. Ter Haar said clients typically need six to 12 months before they’ll commit to a subscription. That runway matters more than it might seem. During that period Monks is building the system, generating performance data and, ultimately, establishing what good looks like. By the time the subscription conversation happens there’s evidence on the table. Renewal becomes an evidence conversation rather than a pitch. And crucially, that data accumulation also tells Monks how much the system actually costs to run, which makes the subscription pricing far more defensible. It’s a structure that could evolve toward genuine outcome-based pay more naturally than hourly billing ever could.

The case against subscriptions 

It’s still a retainer 

The loudest criticism — and the hardest for agencies like Monks to shake — is that subscriptions are just retainers with better branding. And the case for them doesn’t entirely escape that charge. Cost absorption is a real problem, worth solving. But hiding complexity behind a fixed fee is also, more or less, what retainers have always done.

The arbitrage problem

The more serious concern is what that fixed fee conceals. When an agency manages AI on a client’s behalf behind a subscription there’s an obvious temptation: run cheaper models, keep the margin and the client never knows. Webster calls this “arbitraging tokens” — or profiting from the gap between the AI costs to run and what the client pays. It’s the same dynamic that caused so much trouble with agency trading desks a decade ago, where agencies quietly profited from media buys clients assumed were made purely in their interests. The enterprise token deal actually sharpens that risk. If Monks is securing deep volume discounts and running client work on cheaper infrastructure than the subscription implies, the margin opportunity is significant — and largely invisible to the client.

The accountability gap

This connects to a bigger issue: clients need visibility into how their AI works, not just a monthly bill and a promise. Webster’s argument is that agencies have historically made money from knowing more than clients — about media pricing, audience data and where the margins are. Abstracting AI costs into a subscription fee risks deepening that information gap rather than closing it. And without agreed metrics baked into the contract, the question of whether any of it is actually working is easy to quietly avoid.

The timing problem 

Then there’s the awkward reality of where the tech actually is. Ter Haar himself said 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. The infrastructure story only fully works when the AI is genuinely doing the heavy lifting. Charging a subscription premium before that’s true is getting ahead of what the model can honestly justify. 

Done well, the Monks model is a genuinely new way of running an agency: build always-on infrastructure, absorb the variable cost risk through enterprise compute deals, wrap it in a predictable fee, and use the performance data they accumulate to make renewal easier than re-pitching. Done badly, it’s a retainer with a better story and a new margin-hiding mechanism. The industry hasn’t figured out which version it’s going to be yet — and that’s precisely what most of this week’s criticism was really about.

The business behind streamers’ creator push

Digiday caught up with Alphonse Lordo, partner at Content Partners Capital, the private credit arm of Content Partner to understand the shifting investment case for the creator economy. Here’s our conversation, lightly edited for clarity.

Are the economics of the creator economy finally real?

Lordo: “I think the dollars are real. In the MCN phase five to seven years ago, the dollars were not there. Now these channels make pretty substantial EBITDA. Fast forward seven years and it’s been a one-way ride with YouTube in terms of eyeballs and viewership, and that doesn’t seem like it’s going to slow.”

Is this a pivotal moment for creators — studio builders or another MCN cycle?

Lordo: “We’re at a pivotal moment. The question will be, ‘will this be like the MCNs and just fizzle out — granted, they’re making money — or will all these creators who own these businesses build real leadership teams around them and actually go build a real studio?’ They have a golden moment to really become more than what they are. Some are surrounding themselves with good people. They’re going to have to bring in more traditional executives who can help them navigate growth and their initiatives. And they’ve got to diversify. That’s just my view.”

The durability problem: virality versus asset value

Lordo: “A lot of the successful stuff is very of-the-moment content. We like to invest in content that has long term viability. There’s an established model of monetization in perpetuity for film and TV assets. There’s not necessarily that same model for the poppy, of-the-moment stuff. After you see it once, do you need to see it again? Maybe some content will have perpetuity value, and we’re seeing certain genres have a level of perpetuity value. But the poppy stuff — after you see it once, do you need to see it again?”

Numbers to know

16%: Percentage of B2B CMOs who said they have a plan in place to protect their company’s reputation if a crisis emerges

30.2%: Percentage of creators who said they find brand deals by pitching themselves to potential brand partners

42%: Percentage of U.K. and U.S. B2B CMOs that say declining performance from traditional search has driven them to adapt to GEO and zero-click search

$2.9 billion: Total revenue generated by The Trade Desk for 2025, per the company’s earnings results

What we’ve covered

As it ramps up push to fund AI bets, Meta makes a new play for agencies

While Meta continues to push forward with its AI ambitions, it seems it’s not trying to cut out agencies altogether. In fact, the company has been pitching a new team, dubbed the Agency Growth Collective, to help independent agencies succeed on the platform.

Gary Vaynerchuk built his name telling CMOs they were wrong. Now he’s telling their CFOs instead

During VaynerMedia’s inaugural Modern Marketing Leaders Summit 2026 that took place in London on Feb. 23, CEO Gary Vaynerchuk made the point that while he built his empire selling to CMOs, CFOs are now his most receptive audience.

The numbers behind the WPP whistleblower case tell a story of their own

Buried in the court filings of the lawsuit between former GroupM exec Richard Foster and WPP, is a set of numbers that tell the story of how big GroupM was (and how it allegedly made money)

The fight over principal media has moved on but advertisers are still catching up

The standout number in the WPP whistleblower lawsuit is the 97.4% share of GroupM’s proprietary inventory that its own largest clients weren’t allegedly using.

What we’re reading

Amazon’s $50 Billion Investment in OpenAI Could Hinge on IPO, AGI

The investment would see Amazon initially invest $15 billion, while the remaining $35 billion would depend on whether OpenAI goes public or its AI reaches a level that is generally considered on par with human abilities, according to The Information.

Musk’s xAI and Pentagon reach deal to use Grok in classified systems

The agreement allows the military to use xAI’s model Grok, in classified systems, after Anthropic refused the Pentagon’s demand that the company makes its chatbot Claude available for “all lawful purposes”, according to Axios.

Why TV’s Traditional Sellers Aren’t Going Mad For Microdrama

While China powers 83% of the market, traditional TV sellers see thin margins, ultra-fast cycles and little need for middlemen, according to Deadline.

Meta and AMD Agree to AI Chips Deal Worth More Than $100 Billion

Meta agreed to buy six gigawatts’ worth of AI computing power from Advanced Micro Devices, to power data centers over the next five years, according to Wall Street Journal.