What's an ad agency's fee structure for once media buying and creative optimization run on autopilot?

The shift

The two things an agency historically metered its fees against — hours spent producing and optimizing creative, and a percentage of the media it placed — both collapse toward zero marginal cost. Media buying's optimization layer was already automated by platform algorithms; what's new is that creative production, variant generation, and the optimization loop between them now run largely without billable human hours, so the two dominant fee bases (FTE-hours and media commission) are metering work that no longer scales with value.

The axioms

• An agency's output — finished creative, variants, optimized media placements — scales with billable human hours, so hours are a fair proxy for value delivered.
• Media buying at scale requires the agency's desk, relationships, and manual optimization, so a percentage of media spend is a fair way to price that leverage.
• The agency is the execution engine: clients pay it because producing and running campaigns in-house is more expensive and slower than outsourcing.
• Brand strategy and the "big idea" require senior creative and strategic judgment that is scarce and hard to hire.
• A trusted client relationship — the standing to be in the room, to be believed, to be handed the brand — is built over years and can't be bought instantly.
• Someone must be accountable for the brand's outcomes: legally, reputationally, and for the results the CMO reports upward.
• Taste — knowing which idea is worth making out of a thousand possible ones — is a scarce human faculty, not a production output.

Invalid axioms

1. Output scales with billable human hours, so hours are a fair fee base. Creative production and the create-test-optimize loop now run at near-zero marginal cost and near-unlimited volume; hours no longer track the value delivered. Habit-trap: agencies still price on FTE retainers, timesheets, and "scope" measured in people-weeks, so the more efficient the AI stack, the less the agency can bill for the same result — a direct incentive to hide efficiency or pad scope.
2. A percentage of media spend is a fair way to price buying leverage. The optimization and allocation the commission notionally paid for is done by the platform's algorithm (Advantage+, Performance Max) and increasingly by the client's own AI stack; the agency's manual desk work is no longer what moves the media outcome. Habit-trap: agencies still take 5-15% of media as if manual buying judgment were the value, when the value has moved to strategy and the platforms captured the optimization.
3. The agency is the cheaper, faster execution engine. For a large share of production and optimization, an in-house team with the same models is now as fast and cheaper, because the execution labor the agency marked up has largely disappeared. Habit-trap: agencies still pitch "we can produce more, faster" as the core value proposition — exactly the thing that is no longer scarce or defensible.

Unchanged axioms

1. Someone must be accountable for the brand's outcomes. Legal exposure (claims substantiation, disclosure, privacy), reputational fallout, and the results a CMO stakes their job on still land on a named party. A model can generate the campaign but can't be liable for it, and a client will still pay for an accountable partner who owns the call — this is a fee base that survives when hours don't.
2. The trusted client relationship carries standing that can't be generated. Being handed the brand, being believed in the room, having the CMO route the hard decision to you — that is earned trust, and it stays scarce precisely because everyone now has the same production tools. The relationship, not the output, becomes the thing worth paying for.
3. Taste and big-idea origination reward judgment over production. Picking the one positioning or idea worth making out of a thousand plausible AI-generated ones is a human call under real uncertainty, and models are structurally biased toward the median of their training data. The scarce act is deciding what's worth saying, not saying it at volume.
4. Brand strategy under genuine ambiguity is judgment, not synthesis. Category-defining bets, repositioning under competitive threat, deciding what a brand should stand for over a decade — these are high-stakes, low-pattern decisions where being confidently plausible is not the same as being right. This is where senior strategic judgment stays scarce and chargeable.

New axioms

1. What fee base replaces hours and media percentage when execution is automated. If value no longer correlates with either labor or media volume, the industry needs a pricing model tied to what stays scarce — outcomes, accountability, strategic judgment, standing. Value-based and outcome-based pricing have been discussed for decades and never displaced the hourly model; the difference now is that the hourly base is actively eroding, which forces the question rather than merely inviting it. Unresolved: outcome pricing exposes the agency to results it only partly controls, and no one has a clean formula for pricing "accountable judgment" per se.
2. Whether the agency repositions as accountable strategist or dies as an execution shop. The execution-shop half of the business is being commoditized from both sides — platforms above, in-house AI teams below. The open problem is whether an agency can credibly reprice around the scarce half (strategy, taste, accountability, relationship) fast enough, or whether that half is too small a revenue base to sustain the current cost structure.
3. What stops clients from bringing it all in-house with the same tools. The agency's production and optimization moat was labor cost and tooling access; both are gone. If the only remaining reasons to keep an external partner are outside perspective, accountability, and senior taste, the client will pay for a much smaller, more senior engagement — and the industry has to solve for a business that bills far fewer hours at far higher stakes.
4. How to price when the audience for the ad is increasingly an AI agent, not a person. As shopping and research agents mediate more purchases, the optimization target shifts, and the agency's remaining strategic value may hinge on understanding machine-mediated persuasion — an area moving fast and not yet a settled discipline.

Where it breaks

"Hours are a fair fee base" (invalid) collides directly with "what replaces hours when execution is automated" (new): every efficiency the agency's AI stack delivers reduces billable hours, so under the current model the agency is financially punished for the productivity it's selling — and until a new fee base exists, the rational move is to suppress efficiency, which is exactly the behavior a client with the same tools in-house will detect and defect from.

"The agency is the cheaper, faster execution engine" (invalid) collides with "whether the agency repositions as accountable strategist or dies as an execution shop" (new): the pitch that still wins most reviews — more output, faster — is a pitch for the commoditized half of the business, so agencies are actively marketing the part that clients can now do themselves while under-pricing the accountable-judgment half that's the only durable fee base left.

ai-axioms.vercel.app/axioms/marketing/advertising-agency-fees-ai-axioms