Agency utilization rate is billable hours divided by available hours, expressed as a percentage, so a designer who bills 30 of 40 available hours in a week runs at 75 percent. Across the industry, SPI Research’s 2026 Professional Services Maturity Benchmark puts average billable utilization at 66.4 percent for 2025, a record low, with high-performing firms at 75.0 percent. Healthy targets for small agencies sit between 65 and 80 percent. The metric still works for staffing, and it no longer tells you what production costs.
How to calculate agency utilization rate
The formula is simple: utilization equals billable hours divided by available hours, times 100. The arguments start with the denominator. A full-time year is 2,080 hours at 40 hours a week for 52 weeks, and subtracting 15 days of PTO and 10 holidays, 200 hours, leaves 1,880 hours of available capacity.
Some agencies use 2,080 and call the result gross utilization. Some subtract PTO, holidays, and training and call it net. Pick one and use it everywhere, because a 75 percent gross figure and a 75 percent net figure describe different weeks.
The numerator has the same problem. Billable means hours a client can be invoiced for, not hours logged to a client project. An hour of unbilled scope creep is worked, not billable. TMetric’s analysis of 250 agencies cites 23 percent of billable time never reaching an invoice, which is the gap between utilization on the timesheet and utilization on the revenue line.
Worked examples for a five-person agency
Here’s the calculation for an illustrative five-person team on a 40-hour week. The figures are round numbers chosen to show the math, not a real payroll.
| Role | Available hours/week | Billable hours/week | Utilization |
|---|---|---|---|
| Designer | 40 | 30 | 75% |
| Developer A | 40 | 32 | 80% |
| Developer B | 40 | 28 | 70% |
| Project manager | 40 | 20 | 50% |
| Founder | 40 | 8 | 20% |
| Team | 200 | 118 | 59% |
Annualized, that team has 5 times 1,880, or 9,400 available hours. At 59 percent, that’s 5,546 billable hours, and at a $165 blended rate the ceiling on hourly revenue is $915,090 for the year. Every percentage point of team utilization is worth 94 hours, or $15,510 at that rate. That’s why agencies chase the number: the arithmetic is immediate and the levers look obvious.
Two things stand out. The founder at 20 percent and the project manager at 50 percent drag the team average to 59 percent even though the three producers average 75 percent. And the revenue ceiling is fixed by hours, so if those three producers get faster, the ceiling doesn’t rise. It falls.
What is a good utilization rate for an agency
SPI Research’s 2026 benchmark, drawn from 509 professional services organizations with $63 billion in combined services revenue, provides the strongest public comparison. Billable utilization averaged 66.4 percent in 2025, below the previous record low set in 2024. High-performing organizations reached 75.0 percent compared with 64.9 percent for everyone else, while firms at the top maturity level reached 81.2 percent.
Agency-specific guidance runs in the same band. TMetric’s compilation puts agencies of one to ten people at 70 to 85 percent, eleven to fifty at 60 to 75 percent, and fifty-plus at 65 to 70 percent. Producers should run higher than the agency average, because the agency average includes account, project, and leadership time that was never meant to be billed.
Utilization also can’t go to 100. Above about 85 percent for an individual, the hours that disappear are the ones spent on estimates, training, internal systems, and the specification work that makes the next project faster. We’ve watched that happen at our own agency: building a specification library takes hours that can’t be billed to a client, and it has to happen during the same week the team is expected to deliver.
Why AI breaks the utilization metric
Utilization measures hours sold. Hours are no longer what production costs, and that’s the break. Take the landing page from the book: 28 hours of work now takes nine, so at $165 an hour the invoice drops from $4,620 to $1,485.
The designer who did it is at the same 75 percent utilization as before. Her output tripled, revenue per page fell 68 percent, and the metric didn’t move.
The same thing shows up at every scale. A twenty-hour development module done in four hours at $175 bills $700 instead of $3,500. Compress a two-hour task into thirty minutes at the same rate and you’ve removed three quarters of the revenue attached to it.
On one twelve-month time-and-materials engagement at our agency, the first milestone came in about 75 percent under budget. The project was in great shape. The contract wasn’t.
Meanwhile the real cost of production moved somewhere the timesheet can’t see. Our code review system reviewed 1,164 pull requests, ran 2,812 reviews, and posted more than 14,500 findings over 148 days, for $161.20 in total AI spend. That’s about 14 cents per pull request, and no billable hour describes it.
Anthropic’s figures put agents at roughly four times the tokens of chat use and multi-agent systems at roughly fifteen times, so compute is now a real cost of goods, and it scales with output, not with hours.
Utilization can even improve while the business gets worse. If a workflow doubles output and the team spends the freed hours on more hourly work at the same rate, utilization holds and revenue per deliverable halves. Productive’s 2025 survey of 180 agencies found 27 percent had already been asked to cut prices because of AI. The clients can see the hours falling too.
What to watch instead of utilization
Utilization still answers one question well: do we have the right people for the work in front of us? For the economics, we’ve moved to numbers that follow output rather than time.
Revenue per employee is the first. Promethean Research’s 2026 survey of 119 digital agencies found an average of about $163,000, top performers near $250,000, and firms below roughly $120,000 showing deeper structural problems. SPI’s 2026 benchmark put the services-wide average at $168,000, up 6 percent, and revenue per billable consultant at $210,000. Digital Applied’s Q1 2026 survey found the agencies with the strongest AI returns, its “AI-native” 12 percent, ran four to eight production agents and priced on throughput rather than hours.
Margin per engagement is the second. SPI found project margins at a five-year high of 37.7 percent in 2025, with time-and-materials at 36.4 percent and fixed price at 37.2 percent. Promethean’s agencies averaged 35 percent project margin, though only 59 percent of them tracked it. If you track one number per project, track this one, with compute inside the cost.
The third set is what tells you the system is compounding. We watch the hours required for the same kind of deliverable across engagements, the revision rate on first drafts, how much of a new engagement starts from something we already have, and whether the exception queue shrinks as output grows. If the fifth site build in a category takes the same effort as the first, utilization can look fine and the agency hasn’t learned anything.
| Metric | What it measures | 2026 benchmark | Source |
|---|---|---|---|
| Billable utilization | Hours sold / hours available | 66.4% avg, 75.0% high performers | SPI Research 2026 |
| Revenue per employee | Output per person | $163K avg, $250K top, <$120K at risk | Promethean 2026 |
| Revenue per employee | Services-wide | $168K, up 6% | SPI Research 2026 |
| Project margin | Margin per engagement | 37.7% avg, 45.1% high performers on T&M | SPI Research 2026 |
| Net margin, under 10 staff | Bottom line | 19% | Promethean 2026 |
| AI return on spend | Output per compute dollar | 3.2x median, 0.7x bottom quartile | Digital Applied 2026 |
What utilization means for an agency founder now
Keep calculating utilization, because it’s the right tool for deciding whether to hire, and a producer stuck at 40 percent is still a problem you need to see. Stop running the business on it. The moment a workflow compresses twenty hours into four, the metric rewards you for keeping the slow version. Revenue per employee, margin per engagement with compute counted, and the hours-per-deliverable trend across projects are what tell you whether the agency is getting better or just busier.
The Cognitive Agency walks through how we made that switch at our own agency, including the pricing conversations that came with it. If you’re also reconsidering the delivery model behind those numbers, our article on the AI-native software development lifecycle shows what the operational change looks like in practice. Our AI consulting and development team can also help you model it against your own work.
