What happens to agency pricing when AI does the labor?
The short answer
Agency fees have historically priced labor hours, and published rate guides still show the familiar bands: roughly 10 to 20 percent of ad spend, or monthly retainers from about $1,500 to $10,000 and up. As AI systems compress the hours underneath those fees, labor-priced structures face pressure while outcome-aligned pricing holds its logic. The compression hasn't reached published rate cards yet, so it's showing up first in what a given fee buys rather than in the fee itself.
Every common agency pricing model is, underneath, a price on human hours. A percentage of spend approximates how much labor a bigger account needs. A retainer prices a bundle of recurring labor. Even project fees are hour estimates wearing a fixed number. That worked for as long as the labor was the constraint.
Systems change the constraint. When monitoring runs on a schedule, reports draft themselves for a human to refine, and analysis that took an afternoon takes a prompt, the hours underneath the fee compress… and the pricing models built on those hours start pointing at the wrong thing.
The three models under compression
| Model | What it priced | What happens as systems absorb the labor |
|---|---|---|
| Percentage of spend | Labor scaling with account size | The labor no longer scales with spend, so the fee’s logic weakens even where the fee survives |
| Flat retainer | A recurring bundle of hours | The same fee buys either more work or more margin, depending entirely on whether the agency passes the compression on |
| Outcome-aligned (rev share, performance fees) | Results, not hours | The logic holds. Compression changes the agency’s cost, not the client’s basis for paying |
The fork nobody has resolved
Here’s the honest part: published rate guides haven’t visibly moved. Percentage-of-spend bands and retainer ranges look roughly like they did before the tooling wave. Two futures are both live, and we kind of expect the market to split between them:
- Compression gets passed on. Agencies with systems price below the market for the same involvement, and labor-priced fees ratchet down as buyers catch on.
- Compression gets pocketed. Fees hold, agencies with systems run fatter margins or reinvest them, and the visible price never moves… the value shift hides inside scope.
Which future you get as a buyer probably depends less on the market and more on the specific agency’s incentives, which is why the useful question in a pitch has changed.
The question that replaces “how much”
“What do your fees buy in hours” used to be a reasonable probe. It’s becoming a worse one, because hours are exactly the thing compressing. The sharper probes:
- What runs as a system versus by hand? An agency that arrives with monitoring, reporting, and analysis infrastructure spends its human hours on judgment. One that arrives with only labor bills you for assembling.
- Where does the compression go? Fair question to ask directly. Lower fee, more scope, or more margin are all coherent answers, but the agency should have one.
- Is any part of the fee tied to outcomes? Alignment survives compression. Labor pricing doesn’t, at least in logic.
What we don’t know yet
The timeline. Pricing norms move slower than tooling, since buyers anchor on published ranges and agencies have no incentive to lead a race down. It took years for media commissions to give way to retainers; this shift could take as long, or a downturn could accelerate it in a quarter. We’d also flag the possibility that labor pricing persists indefinitely for the relationship half of the work… the judgment hours that don’t compress.
Where we land
We took a position on this rather than waiting: our engagements run as a base retainer plus a revenue share, so part of what we earn only shows up when results do. Systems compress our hours, and the revenue share means the compression’s upside is shared rather than hidden in margin. Today’s market models are mapped in our pricing guide; this page is the bet on where they go.