Marketing agency profit margins averaged about 13 percent after-tax net in 2025, according to Promethean Research’s 2026 State of Digital Services, compared with a long-term average near 15 percent. Agencies with fewer than ten people averaged 19 percent. Agencies with more than fifty averaged 8 percent. AI didn’t create that margin pressure, but it has exposed a weakness in the way many agencies price their work: when revenue is tied to time, getting faster can mean getting paid less.
The month output went up and revenue went down
Picture a month when your team ships more work than it has all year. Turnaround is faster, revisions are down, and two clients call it the best work they’ve seen in years. Then you reach the revenue line and find that it’s down 14 percent. That’s the efficiency trap. The team produced more value, but the business captured less of it.
A landing page that took 28 hours now takes nine. At $165 an hour, that’s $1,485 instead of $4,620. A logo concept that took six hours of sketching now takes about two hours across Midjourney, refinement, and presentation prep, so at $185 an hour the invoice drops from $1,110 to $401. A development module that took twenty hours takes four, and at $175 an hour that’s $700 instead of $3,500.
Compress a two-hour task into thirty minutes at the same hourly rate and you’ve removed three quarters of the revenue attached to it. At our own agency, when we looked at our most productive quarter on record, we worked out that billing the same body of work the year before would have produced two or three times the revenue. Nothing about the work got worse. The pricing model just paid us less for it.
What agency profit margins look like in 2026
Promethean surveyed 119 digital agency owners and managers in February 2026, 68 percent of them founders or partners and 74 percent US-based. Average revenue per employee came in at about $163,000, with top performers near $250,000 and firms below roughly $120,000 flagged as structurally vulnerable. The same survey found the industry’s average after-tax net margin at 13 percent for 2025, down from a historical average of about 15 percent.
Two other 2026 data points help frame the range. SPI Research’s 2026 Professional Services Maturity Benchmark, covering 509 services organizations, found project margins at a five-year high of 37.7 percent while billable utilization fell to a record low of 66.4 percent. Promethean found that agencies with a narrower service focus averaged 13 percent growth and 30 percent net margins, more than double the industry average.
| Benchmark | Figure | Source |
|---|---|---|
| Industry average net margin, 2025 | 13% | Promethean Research 2026 |
| Historical average net margin | ~15% | Promethean Research 2026 |
| Net margin, 0–9 employees | 19% | Promethean Research 2026 |
| Net margin, 10–24 employees | 12% | Promethean Research 2026 |
| Net margin, 25–49 employees | 9% | Promethean Research 2026 |
| Net margin, 50+ employees | 8% | Promethean Research 2026 |
| Net margin, narrowed service focus | 30% | Promethean Research 2026 |
| “Golden era” operating margin | ~30% | VoxComm and Lodestar 2025 |
| Current operating margin, same report | ~10% | VoxComm and Lodestar 2025 |
| Project margin, all services firms | 37.7% | SPI Research 2026 |
| Project margin, T&M vs fixed price | 36.4% vs 37.2% | SPI Research 2026 |
The two headline sources don’t measure the same thing. Promethean reports after-tax net margin from a digital agency survey, while VoxComm and Lodestar report operating margin across a broader agency population. Read them for direction, and the direction is down.
Why margins fall as headcount grows
The Promethean headcount bands are the most useful numbers in the set because they show a mechanism, not just a level. Firms with fewer than ten employees averaged 19 percent net margins. Ten to twenty-four averaged 12 percent, twenty-five to forty-nine averaged 9 percent, and fifty and up averaged 8 percent. Profitability declined at every step.
That isn’t because big agencies do worse work. Every new person adds capacity, but they also enter an organization that has to coordinate, manage, and support them. More projects mean more producers, and more specialists create more dependencies. Growth in the traditional model was linear, and every layer added coordination cost before it added revenue.
Scale still buys real things. Promethean found large agencies grew 2.7 times faster than small ones in 2025, and they hold procurement relationships, roster positions, and the balance sheet to absorb a bad quarter. What the margin data says is that adding people is a harder way to improve profitability the larger you already are.
Why AI efficiency cut margins first
Two things happened to agencies at once. Production got faster, and the pricing model kept paying for hours. One twelve-month time-and-materials engagement at our agency, worth roughly a million dollars in potential revenue, came in about 75 percent under budget at the first milestone. That was a great outcome for the client and a clear problem for the contract.
Clients notice the same gap from their side. Productive’s 2025 survey of 180 agencies found 27 percent had already been asked to lower prices because of AI, and nearly half expected the request. Its June 2026 follow-up of 174 agencies found that discount requests hadn’t increased in the six months since, but 61 percent of agencies still implementing AI were working out their pricing model, up from 47 percent.
The efficiency itself can also fail to show up as margin. Digital Applied’s Q1 2026 survey of 250 agencies found a median return of 3.2 times AI spend, but the bottom quartile recovered only 70 cents per dollar at a median cost of about $1,800 per agent per month. The bottom-quartile agencies usually had no reliable way to evaluate whether the output was better, faster, or worth anything at all. We’ve had workflows that looked efficient until we counted the review and correction time.
Deloitte’s 2026 State of AI report found 74 percent of organizations want AI to grow revenue while 20 percent say they’ve achieved it. That gap isn’t a technology problem. You can get much more productive without changing how the business makes money, and we did exactly that for a while.
What margin structure looks like when compute is a cost of goods
The margin math changes once you stop treating AI as overhead and start treating it as a delivery cost. Token pricing, API metering, and model inference are the cost of goods for an orchestrated agency. Anthropic’s own numbers put agents at roughly four times the tokens of ordinary chat use and multi-agent systems at roughly fifteen times, so a multi-agent run costs close to four times a single-agent one.
At current list prices, that cost is small per unit and potentially large in aggregate. Anthropic lists Claude Opus 5 at $5 per million input tokens and $25 per million output tokens, Sonnet 5 at $2 and $10, and Haiku 4.5 at $1 and $5. Google lists its third-generation Pro model at $2 per million input tokens and $12 per million output tokens for prompts up to 200,000 tokens. Our code review system reviewed 1,164 pull requests across 148 days for a total of $161.20 in AI spend. A production system built on several agents costs considerably more, and Digital Applied’s $1,800-per-agent median is a better planning number than our $161.
We treat compute the way we treat freelancer costs or software licenses: as a line inside delivery alongside people. An agency running this way has three cost bases to watch: people, compute, and the unbilled time spent building specifications and review criteria. KPMG’s Q2 2026 Global AI Pulse found that organizations with full visibility into AI operating costs were five times more likely to report established ROI, 15 percent compared with 3 percent. If you can’t see the compute line, you can’t price it.
What agency profitability means for a founder now
The benchmarks say a small agency can still run at 19 percent net while a large one runs at 8 percent, and they say the average dropped from 15 to 13. Neither number is a ceiling. The agencies posting 30 percent margins narrowed what they sell and stopped passing every efficiency gain straight to the client as fewer hours. That’s the margin play now: change what the invoice is for, then meter what it costs to deliver.
I’ve written up how we worked through this at our own agency, including the ninety-day pricing transition and the mistakes, in The Cognitive Agency. If you’re trying to understand where AI can improve your operation without eroding the value of the work, our AI consulting and development team can help you model it.
