Part of a series from the Women in Marketing Procurement Summer Series session “What AI Really Does to Your Content Budget,” featuring The Brandtech Group’s Karen Bennett (Jellyfish), Matt Franceschi (Pencil), and Tamara Carvalho (Oliver).
——————————————————————————-
Most of us in marketing procurement have a creative services rate card somewhere in a shared drive. It lists agency roles, hourly rates, and overhead multipliers, and it was negotiated carefully and in good faith. It has served us well for years.
Here is the question our recent Women in Marketing Procurement session left us with: is that rate card still describing the work our agencies actually do?
The session with The Brandtech Group was about taking a look at what gen AI does to the creative content budget, and it delivered.
However, the bigger story for marketing procurement sits one level deeper. When AI platforms can produce content at ten times the volume in a fraction of the hours, compensation models built on time and headcount start to lose their logic. That is not a criticism of anyone. It is simply what happens when the economics underneath a pricing model change.
The good news? The session also showed us what the alternatives look like, in live contracts at some of the biggest advertisers in the world. Let’s walk through them.
The Diageo Signal
The example that caught our attention was a major liquor company. As Matt Franceschi, who leads on Brandtech’s Pencil platform, described it, Pencil is sold to the brand as a central piece of technology that the company owns and operates. The brand can then invite its partners in. This company took that logic all the way: the company asked its roster of 30 to 35 global agencies to work inside the client-owned platform.
Think about what that changes. For decades, the technology, the workflows, and the data lived on the agency side. The client bought hours and trusted that outcomes would follow. Now the client owns the operating system, and the agencies work within it. As Franceschi explained, the goal is “accruing the benefit, speed, cost, and effectiveness, accruing that benefit to you centrally”, rather than leaving it with any single partner.
The company has been open about this journey. Speaking to Campaign about the company’s virtual content studio, which includes Pencil among its AI tools, Diageo’s Gemma Jones acknowledged that AI has changed the company’s agency relationships and framed it constructively: it is about identifying which activities no longer add value, so both sides can focus on the ones that do. The creativity stays sacred. The production economics evolve.
For procurement, the question follows naturally. If our agencies are producing far more work in far fewer hours, what exactly is an FTE-based fee paying for? That is not a gotcha. It is the honest starting point for the next negotiation, and the agencies are asking themselves the same thing.
Why the Hourly Model Is Under Pressure
Labor-based fees were always a proxy. We could not easily measure the value of agency work, so we measured the cost of producing it and added a margin. That worked reasonably well as long as effort and output moved together.
AI loosens that link. Franceschi shared benchmarks drawn from work with three of the world’s ten largest advertisers, built on the assessment of nearly a quarter of a million pieces of content, and the consistent pattern was faster, cheaper, and better. Brandtech points to one client producing content at ten times previous volume through its Pencil-powered studios, and to another client pairing a meaningful ROAS improvement with a 33% cost reduction. As always, these are vendor figures and deserve verification. But even at half the size, the direction is clear: output is growing while hours shrink.
Here is the part worth sitting with, because it affects our agency partners as much as us. Under a purely hourly model, an agency that invests in AI to serve us better reduces its own revenue with every efficiency it delivers. That is not a healthy incentive for anyone. If we want our partners to embrace these tools, the commercial model must reward them for doing so, not punish them.
The industry data shows how much room we have to improve. The ANA’s Trends in Agency Compensation research found in its 19th edition that 84% of marketers used fee-based compensation in at least one agency agreement, down one percent from 2022. The same study found the use of performance incentives falling to 15%, with many marketers unsure whether those incentives were actually improving agency performance. The 19th edition arrived in late 2025 and remains the benchmark for anyone renegotiating today. In short: the dominant model prices inputs, and the outcome-based layer is thin. AI gives us a reason, and an opportunity, to fix that.
Three Models Taking Shape
So what replaces hours? Three approaches are emerging, and we saw elements of all three in this one session.
- Output and deliverables-based pricing. You pay for the asset, the campaign, or the localization set, rather than the hours behind it. At an ANA session in February 2026, agency search consultant Joanne Davis made a practical point in favor of this model: it makes AI savings visible. Joanne said “when a master asset plus 40 versions cost X last year and 0.6X this year, everyone can see the progress and talk about it openly”. Under a blended retainer, that same efficiency is much harder to identify and capture.
- Subscription, or “advertising as a service”. Writing in MediaPost in April 2026, Maarten Albarda, who helped develop early value-based compensation models at Coca-Cola, described a model where the client buys outputs, such as a steady flow of social assets or managed search performance, and how the agency delivers them, with people or AI, is the agency’s business. Importantly, Albarda scopes it carefully: this suits repeatable, productized work, while value-based pricing remains the right tool for high-stakes strategic work such as brand repositioning. Hybrid structures, in other words can help increase visibility while still paying for strategic work that is hard to put a dollar value on.
- The tapering hybrid. The most instructive case Franceschi shared was from a major UK retail bank. The engagement was designed as a journey: Pencil technology plus a full managed service from an Oliver team in the early years, because the client knew it was not ready to run alone, with the service component tapering down as internal proficiency grows over a three-to-five-year horizon. “We meet our clients where they are at” as Franceschi put it.
For procurement, the key word is designed. The commercial model anticipated its own evolution. The year-one fee structure was never meant to survive to year four, and everyone signed knowing that. Compare that with how most of us contract today: a model negotiated once, rolled forward annually with an inflation adjustment (sometimes). The above structure treats compensation as a living mechanism, and that is the mindset AI-era contracts will need.
Buy the Pieces, Not the Bundle
One more moment from the session deserves a highlight, because it goes to the heart of commercial flexibility.
Many large agency groups are now arriving at pitches with a proprietary AI operating system, and the offer often comes as a package: the platform plus the group’s agencies, together. We asked the panel directly how Brandtech’s model differs, and the answer was encouraging: you can work with the Pencil technology alone, with Oliver’s in-housing model alone, with Jellyfish as an agency alone, or in any combination, and the platform is open to any other agency you choose to bring in.
For procurement, this points to a simple principle worth applying to any platform, from any provider: contract the pieces separately, even when you buy them together. Separate the technology license from the service fees. Give each its own pricing, its own review points, and its own exit provisions. When the components are priced individually, you can benchmark them individually, and you keep your flexibility as your needs evolve.
Karen Bennett added a scoping insight that belongs in every AI-era statement of work: adoption is a spectrum of comfort. Some brands want only translation and templated versioning. Some are happy to add movement to static images but are not ready for fully generated people. Some want complete 30-second AI-generated films. Each point on that spectrum is a different scope, a different risk profile, and a different price. Defining where your organization sits, market by market and brand by brand, is genuinely useful work, and it is work procurement can lead.
Five Practical Moves for the Next Negotiation
Start with the repeatable work. Versioning, localization, resizing, translation: this is where AI economics are clearest and where output-based pricing is easiest to introduce. Move this work to deliverables-based fees first and let strategy and original creative follow as measurement matures.
Build the taper into the contract. If your engagement includes a managed service alongside technology, follow the “walk first, then run” logic detailed above: agree the glide path by which service fees reduce as internal capability grows, with defined review gates rather than good intentions only.
Ask how AI efficiency flows through to fees. This is a fair, collaborative question, and good partners will welcome it. If the platform produces your work at a fraction of last year’s cost, the pricing conversation should reflect that, and both sides benefit from having it early rather than late. Acknowledge that there is technology investment and look at this as we used to look at agency overhead cost.
Refresh your benchmarks. Every production benchmark we hold, cost per asset, cost per adaptation, hours per campaign, was set in a pre-AI world. Benchmarking new fees against old numbers will mislead us, in both directions. Re-basing them is foundational work for the next cycle.
Protect the agency’s margin. This is not a race to zero, and it should not become one. If we drive fees down in lockstep with hours while expecting the same strategic quality, we weaken the partners we depend on. The mature move is to pay properly, and sometimes pay more, for the judgment, strategy, and creativity that AI cannot replicate, while no longer paying yesterday’s price for what a platform now does in minutes. The win is the reallocation, not the cut.
Procurement’s Opportunity
The ANA’s compensation research noted years ago that procurement’s role in agency compensation was growing steadily. AI brings that trajectory to its natural destination. Compensation design is no longer an annual administrative exercise. It is the mechanism that decides how the value of the AI transition gets shared between client and agency.
That is a genuinely exciting place for our function to be. The fee models of the last thirty years were built around hours because hours were what we could count. Now we can count so much more: outputs, outcomes, speed, and quality. The organizations that redesign their agency compensation thoughtfully, with their partners rather than against them, will capture the value of this transition and keep their agency relationships healthy while doing it.
The compensation model is changing. Our job is to make sure it changes by design.