The average US advertising agency had 10.2 employees in 2015 and 5.5 by 2024 (BLS QCEW). Over those nine years the number of agencies doubled, from 19,078 to 39,052, while total headcount grew 10%. The people stayed in the industry. They redistributed into twice as many, smaller shops.
If you run one of those shops, one planning question keeps coming back. How do you take on more client work than five people can deliver? The reflex answer is a sixth person. The data says that's expensive, and the agencies growing fastest pull a different lever.
Half the size, twice as many
Small is where the industry lives. 84.2% of US advertising and PR businesses have no employees at all: 207,863 solo operations against 38,864 firms with staff (US Census Nonemployer Statistics and County Business Patterns, 2023). Of the agencies that do employ people, 68.2% have fewer than five and 81.0% have fewer than ten (Census County Business Patterns, 2023).
Europe says the same thing with a different instrument. 78.8% of EU advertising enterprises employ zero or one person, up from 76.4% in 2021 (Eurostat). Tax returns in one economy and business registers in another land within six points of each other. The micro-agency is the agency.
It's also the part that's growing. Solo advertising businesses grew 41.3% between 2015 and 2023 while agencies with staff grew 3.6% (US Census Nonemployer Statistics, 2023). Receipts per solo shop rose 10.9% over those eight years, which after inflation is a decline. More small shops are competing for the same money with the same number of hours.
That is the constraint. A five-person agency runs out of build capacity long before it runs out of demand.
Why the sixth hire doesn't pay
Headcount adds cost faster than it adds output. Promethean Research attributes the margin gap between small and large agencies to the management layers, support roles, and coordination that arrive with growth (Promethean Research, 2026). The full margin-by-size numbers are in what agency work actually lasts; the short version is that the gap runs the wrong way for anyone planning to grow by hiring.
Doing less did better. Agencies that reduced their service list grew 13% in 2025 against a 7.5% industry average, and Promethean puts their net margins near 30% versus 10% for agencies that expanded, a figure it calls a directional benchmark (Promethean Research, 2026). Small sample, correlation only, and still the clearest signal in the data.
Pricing didn't rescue anyone either. Value-based pricing fell from 37% of agencies in 2022 to 18% in 2025 (Promethean Research, 2026). Whatever AI did to agency economics, it did not push agencies toward charging for outcomes. Where the gains exist, they show up as capacity: more delivered per person.
What the extra capacity is for
Promethean's report says AI "is creating new demand for strategy, implementation, governance, workflow redesign, automation, and agent development" (Promethean Research, 2026). Forrester's 2026 prediction has agencies "selling execution, managed services, proprietary products, and strategic partnerships" (Forrester, October 2025). Both point the same way. Clients want the thing built. They want the agency to build it.
The people doing the building have changed too. 43.8% of designers now spend more than half their building time on AI-generated code (UX Tools, Spring 2026, n=1,478). Another 31.1% say most or nearly all of their workflow is AI-generated. Designers participating in development work doubled to 41% (Figma 2026 AI Report). The designer who used to hand a Figma file to a subcontractor is becoming the person who ships it.
How agencies deliver client websites faster without adding a developer
The bottleneck in a small agency sits at the step where the design becomes something a client can click. That step used to mean a developer, a queue, and revision rounds measured in tickets.
Bolt.new removes the queue. You describe the deliverable, the audience, and the brand direction, and Bolt builds a working site or app you can see and test in the browser while the client is still on the call. Revisions happen in the meeting instead of the following week. A client portal, a campaign microsite, or a booking system ships as a live URL, and the code stays exportable to hand over when the engagement ends.
The capacity shows up in three places. In the pitch, a running prototype wins scope a deck can't. In the revision round, a change made while the client watches stops costing a week. And the client who came for a website is the one who needs the intake form and the dashboard next, which is the work that tends to keep running after launch (what agency work actually lasts).
The tension worth naming
Some clients will pick up these tools themselves. A product that lets a five-person agency ship a portal in a week lets a client's marketing lead attempt one too. The agencies that keep that client are doing work the client can't self-serve: the strategy, the integration with systems the client already runs, the ten edge cases the demo skipped. That is also where Promethean and Forrester say the demand is heading.
Most of the restructuring happened before AI build tools were mainstream. The small shop won on its own terms, and then the build step got shorter. Five and a half people, and no ticket queue between the design and the deliverable.
Start with the next brief. See what agencies build in-house with Bolt for agencies.
