The model hasn’t changed. The world has.

The agency business model is, at its core, a beautiful arbitrage. You hire smart people. Smart people develop expertise. Expertise gets applied to client problems. Clients pay for that expertise, either as a percentage of media spend or as a fee for time. The agency pockets the difference between what it costs to house the smart people and what clients pay for access to them.

The arithmetic is unusually legible. A well-run media agency earns roughly 10–15% of the media it manages as net revenue — the actual fee it keeps after passing client money through to media owners. At a blended $200–250K in net revenue per person, and with senior planners in Sydney costing $120–180K all-in, the margin in the middle is where the business lives.

This model has worked for decades because the inputs — media knowledge, planning frameworks, trading relationships, campaign intelligence — took years to develop and were genuinely scarce. Junior staff learned from senior staff. Senior staff learned from doing. The institutional knowledge lived in human heads and PowerPoint decks and the collective memory of a team that had planned a hundred campaigns together.

That scarcity is dissolving. And with it, the margin that sits in the middle of the arbitrage.

What AI is actually doing to this

Here’s what’s already happening, right now, in 2026. An AI system can ingest a client brief, access real-time media consumption data, model audience segments, generate a channel strategy, draft a media plan with reach and frequency curves, write the rationale document, and produce a post-campaign attribution model — faster than a junior planner can book the briefing meeting.

Is it perfect? No. Does it replace a 20-year veteran with deep trading relationships and genuine creative instinct? Not yet. But “not yet” is doing a lot of heavy lifting in that sentence.

The more honest framing: AI doesn’t need to replace the best thinking in the room. It just needs to replace the average thinking — the standard channel recommendations, the boilerplate competitive analysis, the cookie-cutter audience segmentation, the templated reporting. And that is the work that fills most of the billing hours across most agencies.

The savvier agencies know this. Some are building proprietary programmatic stacks. Others are acquiring digital measurement capabilities or developing first-party identity solutions. These are all, essentially, bets that owning technology creates defensible margin that pure human capital cannot.

They’re right. But building technology inside an agency isn’t the same as becoming a technology business. And that distinction is the whole game.

The real question: what are you actually selling?

If AI systematically compresses the value of junior-to-mid-tier planning and buying work — and it will — then every agency in this market faces a version of the same strategic question: what is the thing we sell that AI cannot?

The holdcos have already started answering this by acquiring data and technology assets at scale. The independents are answering it through differentiated positioning — owning a niche, a client segment, a methodology, a measurement approach. These are good answers to a real question.

But they’re answers to the wrong question.

The right question isn’t “what can’t AI do?” It’s “what does the client actually need when AI can do most of the work?” And the answer to that is: a trusted specialist who can tell them what to do with the output.

Enter the AI marketing operating system

This is Frontier’s thesis, stated plainly: the next generation of marketing services won’t be built around people who do thinking. It’ll be built around a platform that does the thinking, and humans who guide clients through executing it.

Think of it like this. For decades, enterprise software came with an army of consultants who implemented it, customised it, and told you what to do with it. SAP didn’t eliminate the consulting business — it created a whole new one. The consultants shifted from doing the underlying work to guiding the deployment of a system that did the work better than they ever could manually.

The same transition is coming for marketing. And it arrives in the form of a subscription-based AI marketing operating system — one that handles strategy generation, media planning, campaign execution, performance optimisation, and reporting in a single integrated workflow. What it needs alongside it is not a 94-person agency. It’s a small, expert team of humans whose job is to sit with the client, understand the business context, make the judgment calls the AI can’t, and guide the execution.

Why the numbers make this inevitable

Consider what the typical agency P&L looks like. A mid-sized independent with $200M in billings earns roughly $24M in net revenue at a 12% take rate. At 80 people — a reasonable headcount for that billings volume — roughly 70–75% of that net revenue goes to salaries. The operating margin that’s left over, typically 15–20%, is the whole game. It’s real, but it’s fragile. One bad pitch cycle, one key person departure, one client loss, and the margin compresses fast.

Now run the same thought experiment with an AI marketing OS at the centre. The same $200M in billings, but the analytical and execution layer is handled by the platform. You need fewer people — and the people you do need are specialists guiding the platform, not juniors executing templates. Net revenue stays the same. Headcount drops materially. Margin expands — not incrementally, but structurally.

And here’s the part that makes the economics genuinely exciting: the capability subscription compounds. Every campaign the system runs makes it smarter. Every brand it learns makes the next brief faster. The human team doesn’t scale linearly with billings growth because the platform absorbs the scale. The agency P&L has always been a people equation — more billings means more people. The AI OS breaks that equation permanently.

The four pillars of an AI marketing OS

The two-part model matters as much as the technology itself. The capability subscription — your access to the AI marketing OS — is tied to outcomes and platform usage, not headcount. It scales with your ambitions, not with an agency’s hiring plan. It compounds in value over time as the system learns your brand. And the specialist team that delivers guided execution isn’t billing hours — they’re guiding a platform that does the heavy lifting, which means their time goes entirely to judgment, context, and client partnership rather than production.

The humans aren’t going anywhere. They’re just doing a different job.

Let’s be clear about something, because the “AI will replace everyone” take is both boring and wrong. The best client relationships in this industry are built on human qualities — judgment, context, trust, the ability to walk into a leadership team meeting and say “here’s what the data means and here’s what we’re going to do about it.” None of that goes away.

In fact, in a world where the analytical work is automated, those human qualities become more valuable, not less. When every agency has access to the same AI infrastructure, the differentiator will be the quality of the specialists guiding it. Knowing which output to trust. Knowing when to override the model. Knowing how to translate machine intelligence into a decision a CMO can act on with confidence.

The shift is from “humans who do thinking supported by some tools” to “AI that does thinking, guided by humans who exercise judgment.” That’s a meaningful distinction. It means the marketing function of the future looks less like a 90-person agency and more like a 15-person specialist team with an extraordinarily powerful platform behind them. Same outcomes. Better economics. Faster everything.

And crucially — the specialists get to do the interesting work. No more junior planners building the same reach-and-frequency slide for the forty-seventh time. No more account managers spending Thursdays chasing trafficking confirmations. The platform handles the production. The humans handle the thinking that actually requires humans.

What this means for marketing teams

Here’s who wins most immediately from this shift: the client side.

Marketing teams have spent the last decade caught between two inadequate options. Option one: hire a large agency and spend half your time managing their process, their account team layers, and their quarterly reviews. Option two: build an in-house team and spend the other half trying to attract, retain, and keep up to speed a group of specialists who can earn more somewhere else.

The AI marketing OS changes both calculations. A marketing team that subscribes to a platform capability and engages a specialist team for guided execution gets the best of both worlds — agency-grade thinking and execution depth, without the agency headcount overhead; in-house control and transparency, without the in-house hiring burden.

It also changes what “sophisticated marketing” means. Historically, scale was an advantage — bigger brands could afford bigger agencies with more research, more trading leverage, more specialist resource. An AI system that learns continuously and costs a fraction of a traditional agency fee makes sophisticated, data-driven marketing genuinely accessible to brands that couldn’t previously afford it at this quality level. That’s not a small shift.

So what does Frontier actually do about this?

We built the operating system. Not a collection of AI tools bolted onto a traditional agency model. Not a chatbot that helps someone write a brief faster. An end-to-end AI marketing capability — from strategy generation through to media execution and measurement — designed from the ground up to be guided by specialists rather than operated by an army of planners.

Here’s what the Frontier model looks like in practice. You subscribe to the capability — the platform that handles the thinking, the planning, the execution, and the measurement. It learns your brand. It ingests your performance history. It models your audience. It generates strategy, optimises in real time, and surfaces intelligence that most agencies would charge you a separate analytics retainer to produce. The platform gets smarter with every campaign. Your marketing compounds.

Alongside the platform, you engage Frontier’s specialist team for guided execution. These are experienced marketing operators — not account managers fielding your calls, but genuine specialists who understand the system deeply and who know how to direct it toward your specific objectives. They contextualise the AI’s outputs. They make the judgment calls. They translate machine intelligence into decisions your leadership team can act on. They are the human in the loop — and they’re considerably more valuable in that role than they ever were in the traditional agency model, because they’re freed from production work entirely.

The two elements work together as a single integrated model. The capability subscription gives you access to infrastructure that would cost millions to build internally. The specialist team gives you the expertise to use it properly. Neither one works as well without the other — which is also why the model is genuinely hard to replicate by bolting AI tools onto a traditional agency structure.

The agency business model is about a hundred years old. The underlying exchange — expertise for money — is even older. What’s changing is what expertise means in a world where intelligence is abundant and cheap. In that world, the most valuable thing is not the ability to generate the thinking. It’s the ability to direct a system that thinks, and know what to do with what it surfaces.

That’s what we’re building at Frontier. And we think the timing, right now, in early 2026, is almost exactly right.

Agency billings data and industry statistics sourced from Mumbrella’s Top Indies ranking (March 2026), the IMAA Indie Census (2025), and TrinityP3 research. Financial benchmarks and enterprise value estimates are illustrative only and based on publicly available marketing services sector data. They should not be construed as financial advice.