The headcount shift inside agencies is already happening, and the numbers are not subtle.
The agencies that are just "using AI tools", giving the team a Claude subscription and hoping for the best, are about to find out that a 30-40% efficiency boost doesn't change anything structurally.
The ones rebuilding their delivery model around AI are playing a completely different game.
Agency AI transformation
I was on a call with the founder of a paid media agency managing around 100 client accounts. His team tracks time across each task: reporting, strategy, creative briefing, client comms, all broken down by media buyer.
One junior media buyer who uses AI properly spends 12 minutes on a client report. A senior media buyer who doesn't takes 55 minutes for the same report, the same data, the same output.
That's a 4.5x difference on a single task. And reporting is one of maybe 12 things a media buyer does in a week.
This founder's first instinct was the same one I hear across the agencies I work with: "Let's give the team better tools. Let's automate the routine stuff and free up the media buyers to focus on strategy." That gets you maybe 30-40% more capacity, which is good. But there's a bigger opportunity here.
He said something that shifted my thinking. He asked himself: "What actually is strategy? Why can't AI do strategy better than a media buyer can, if it has all the context and data?"
If AI has access to the client's performance data, their history, their goals, their creative performance and their market context, the strategy a media buyer produces manually is already worse than what a well-built AI agent could generate.
So the role fundamentally changes. They stop being the person who builds campaigns and analyses data. They become the person who manages the relationship, interprets the AI's output for the client, and steps in when something needs a human decision.
A media buyer managing 8-12 accounts today could manage 50. Maybe more. The execution layer is handled by AI agents that have full context on each client.
The role looks a lot more like a customer success manager than a traditional media buyer.
The generational split is real
In this agency, the ambitious juniors are the ones adopting AI naturally. Building with it, experimenting with it, treating it as a co-pilot from day one.
The seniors, not so much. The ones who've been doing it the same way for three or four years are resistant. And I get it. Being told that the way you've worked for four years is about to change completely feels like starting from a blank canvas.
Spencer Stuart's latest research puts it bluntly: 37% of CMOs at the largest companies say their CEOs expect 20%+ cost cuts within two years. That pressure flows directly into how agencies are structured and who they employ.
The agencies that figure this out in the next 12 months will have a structural cost advantage that's very hard to replicate. Same quality of delivery, fundamentally different economics. A pod that needed five people two years ago might need three. The margin profile changes. The pricing flexibility changes. The competitive position changes.
Watch your AI anxiety
I fell into the same AI trap. For the past half year I've been testing several dozen AI tools, building AI agents, buying AI services and hiring AI engineers. Obsidian, OpenClaw, RAG pipelines, databases, Claude agents and so on.
Some of it is useful. A lot of it turned out to be a waste of time for me as a business owner.
We, as founders, have to be straight with ourselves about what this means for how we structure our teams. The roles that exist today in your agency will look completely different in 18 months. Some of them won't exist at all.
The question is whether you're designing for that future or hoping it doesn't arrive.
What does your agency look like with AI built into the operating model?
The Founder Dependency Audit diagnoses where you're still the bottleneck and what an AI-native delivery model would change first. We work through it together.
See where your agency depends on you- Romans