6 min read
July 2026
Is AI cutting agency jobs? What the data actually shows
Overall job numbers look strong but entry-level opportunities are shrinking.
No, not at the rate the headlines suggest. Aggregate employment holds up in firms that use AI, and the pressure lands somewhere narrower: the entry rung.
If you run a small agency you have probably had some version of the thought anyway. If AI does the production work, how many people are actually needed? Here is what is established, what is genuinely contested, and what it means for a team of your size.
What agency leaders expect
The most cited figures come from a Sunup survey of 225 US marketing agency leaders conducted in August 2025. Those numbers are real and they get quoted constantly. Two things about them are usually left out.
What agency leaders said they expected
The first two numbers get quoted constantly. The third one, from the same survey, almost never does. It points the other way.
First, they measure expectation, not outcome. What leaders think will happen to headcount is not the same as what happens to headcount.
Second, three quarters of those same agencies are actively hiring for AI-focused roles blending creative and technical depth. That is not a shrinking workforce. It is a workforce changing shape.
What has actually happened to employment
This is where it gets more interesting, and less tidy.
Economists at the European Central Bank found no overall employment gap between firms using AI and firms not using it. That drew on a survey of around 5,000 firms, published in March 2026.
PwC's 2026 Global AI Jobs Barometer, published that June, found the opposite of shrinkage. Headcount at the most AI-exposed companies grew faster than at the least exposed, 52% against 36% since 2018. PwC notes this covers larger firms with published accounts, so it is not an economy-wide measure.
Neither finding is consistent with a simple displacement story. At the same time, the entry rung really is under pressure.
Job adverts are falling fastest where AI exposure is highest
Not a collapse across the board. A squeeze concentrated on the roles AI touches most, which in an agency means the entry rung. The report itself cautions that the fall cannot be attributed to AI alone.
So both things are true. Aggregate employment is holding up. The bottom rung is getting narrower.
The most plausible reading is that AI makes the juniors who remain more productive while reducing how many a firm needs at entry level. PwC's Barometer supports that. The most AI-exposed junior roles are seven times more likely than the least exposed to demand traditionally senior skills like judgement and leadership.
In the United States, entry-level roles rewritten that way grew 35% between 2019 and 2025, while comparable roles declined 10%. The number of rungs is contracting. The value of a person on one is not.
The finding that cuts against the narrative
In April 2026, Ad Age spoke to 18 agency executives and hiring leaders about junior roles. The headline was not what you would expect from the discourse. Independent shops are leading entry-level placement.
One experiential agency hired 100 people in two months, most of them entry-level. Others are introducing apprenticeships and shadowing programmes aimed at the higher-level skills AI does not cover.
Small and independent agencies are, on this evidence, hiring juniors more readily than the holding companies. Which makes sense. A twelve-person agency does not have a layer of pure execution roles to remove. Everyone there already does several jobs.
Most of the alarming data comes from large US agencies with structures small agencies do not have. Be careful about importing conclusions drawn from a 2,000-person holding company into a team of fifteen.
The risk nobody prices
A study by the agency DBC with the 4As, published in June 2026, made an argument worth taking seriously. Agencies that treat AI as a replacement for junior staff risk eroding the apprenticeship model the industry runs on. The result is a hollowed-out middle management layer within a few years. This is the real cost, and it does not show up in this quarter's numbers.
Senior people are made, not hired. They are made by doing the unglamorous work, getting it wrong, and being corrected by someone more experienced. If AI absorbs all the unglamorous work, that path closes.
The saving is immediate and visible. The gap arrives in 2030 and cannot be closed quickly, because everyone else made the same decision at the same time.
Past hiring pullbacks came with falling demand. This one is happening while the work is still there. That is what makes it different, and it is why the shortfall compounds faster than people expect.
What this means if you run a small agency
The pressure is real, but it is not primarily a headcount question
Forrester's June 2026 research with the 4As found 81% of agencies using generative AI mainly to improve staff productivity. Most still treat AI as a cost centre rather than a revenue driver. Agencies are getting faster and charging the same or less for it. That is a pricing problem before it is a staffing problem.
The apprenticeship question is a decision, not a trend
You can use AI to remove junior work. Or you can use it to move juniors up the value curve faster than before. Both are available. The second is harder in year one and considerably better in year five.
Capability is a team property, not an individual one
This is the part that gets missed most often. When one person learns to use these tools well and nobody else does, almost nothing changes. The knowledge does not travel. It leaves when they do.
That is why AI adoption is a team problem. Not because everyone needs to be an expert. A workflow that only one person can run is not a workflow. It is a dependency.
The honest summary
AI is not eliminating agency jobs at the rate the headlines suggest. The aggregate employment data does not support that story. What it is doing is compressing the entry level, shifting demand toward senior judgement, and pressuring how agencies charge for time.
For a small agency, the more useful question is not how many people you need. It is whether the capability sits with the team or with one person. And whether the people you have are getting more capable, or just busier.
What to do next
If the person who uses AI most here left next month, what would stop working?
Ask this at your next team meeting. If the answer is meaningful, that is the thing to fix, and it is fixable.
ViraOps is built for that. One flat licence, up to five seats included, so the whole team learns rather than one person. €89 a month, or €890 a year, with a 14-day free trial.
Sources
- Sunup, survey of 225 US marketing agency leaders, August 2025
- Lebastard and Sondermann, European Central Bank survey of approximately 5,000 firms, March 2026
- UK Department for Science, Innovation and Technology, AI exposure and job advert analysis
- IDC study commissioned by Deel, entry-level hiring expectations
- Ad Age, interviews with 18 agency executives and hiring leaders, April 2026
- DBC and 4As, study on AI and entry-level roles, June 2026
- Forrester and 4As, The State Of AI Inside US Marketing Agencies, 2026, published 24 June 2026