The Death of the Agency Model

I spent fourteen years inside the advertising agency model. I worked at independent agencies and creative shops — the good ones, the ones that still believed in the work. I watched pitches get won and lost. I saw the best people burn out and leave. I experienced the entire lifecycle of agency life, from the exhilarating highs of a great campaign to the grinding reality of client management, politics, and the relentless pressure to bill more hours. I know this world intimately. And I am telling you: the model is dying.

Not because agencies make bad work — many still make extraordinary work. Not because clients do not need help — they need it more than ever. The model is dying because the economics no longer make sense, the talent is leaving, and a new alternative has emerged that is faster, cheaper, and often better. That alternative is small teams with AI.

The uncomfortable economics

Here is the slide I would put up in a conference presentation, the one that makes everyone in the room uncomfortable: the cost comparison.

A typical agency engagement to build a brand strategy, design system, and digital presence involves a team of eight to fifteen people over six to twelve weeks. A strategy director, a strategist, a creative director, two or three designers, a copywriter, a project manager, a UX designer, and various production specialists. Depending on the agency, this engagement costs the client somewhere between fifty thousand and two hundred thousand pounds. This is not an exaggeration. These are standard rates at mid-tier agencies. At the large holding company agencies, the numbers go higher.

Now here is the uncomfortable comparison. I built eighteen products — complete digital products with strategy, design, content, and functionality — using AI tools. Not over the course of a year. In weeks. Each one. The total cost of tooling for all eighteen projects combined is a fraction of what a single agency engagement costs. One person. AI tools. No team. No overhead. No six-week timeline.

I am not saying this is a perfect comparison. Agency work and solo AI-assisted work are different things. The agency brings collaborative thinking, client management, production at scale, and the credibility of an established brand. But the gap is closing. And for a growing category of work — brand development, digital products, content strategy, campaign creation — the solo practitioner with AI is becoming a genuine alternative to the agency team.

When one person can do in a weekend what used to take a team of ten six weeks, the economics do not just shift. They collapse.

The talent exodus

The agency model has a talent problem, and it has had one for years. The best people leave. This is not a theory — it is observable reality across the industry. The most talented strategists, creative directors, designers, and writers consistently migrate away from agencies towards freelance work, in-house roles, or their own ventures.

The reasons are well documented but worth restating. Agency hours are brutal. The pressure to bill is relentless. The politics of client management drain creative energy. The hierarchy slows decision-making. The holding company structure prioritises margin over work quality. And the compensation, relative to the hours worked, is often poor compared to alternatives.

But there is a deeper structural problem that AI has now made visible: the agency model requires people to work in ways that are fundamentally inefficient. A strategist spends half their time in meetings, a quarter of their time on administration, and perhaps a quarter of their time on actual strategic thinking. A designer spends hours in revisions that exist because the feedback loop is too slow. A copywriter produces twenty headlines so the client can choose one. The waste is staggering — not because anyone is lazy, but because the model requires consensus, collaboration, and client management at every step.

AI does not need consensus. AI does not need a meeting. AI does not need to be managed. AI produces options instantly and iterates in real time. For the talented individual who has spent years building the judgment to know what is good, AI removes all the friction between having an idea and executing it. Why would that person stay in a model built around friction?

The holding company trap

The large holding companies — the ones that own dozens of agency brands under a single corporate umbrella — face an even deeper version of this problem. The holding company model is built on scale. Its value proposition to clients is: we have people everywhere, in every discipline, and we can coordinate complex, multi-channel campaigns across markets. Scale is the product.

But AI eliminates the need for scale in most of the areas where holding companies sell it. You do not need a production team of twenty when AI can generate assets. You do not need a localisation team when AI can translate and adapt. You do not need a data analytics team when AI can process and interpret data. The functions that justified the scale are being automated, one by one.

What remains is taste, strategy, and creative vision — the things that AI cannot do. And these are precisely the things that do not require scale. They require talent. Small amounts of exceptional talent. The holding company model is optimised for deploying mediocre talent at scale. It is structurally incapable of competing with exceptional talent deployed at speed.

This is not an argument that all holding company agencies will disappear. Some clients need scale — global brands with operations in forty markets need coordination that a solo practitioner cannot provide. But the proportion of work that genuinely requires holding company scale is much smaller than the holding companies would like to admit. And it is shrinking every time the AI tools improve.

What replaces it

The future of creative services is not one model. It is many. And that plurality is itself a departure from the agency era, where "the agency" was the default structure for almost all creative work.

The first model is the studio. A small team — three to eight people — with a specific point of view and a specific area of expertise. The studio does not try to do everything. It does one or two things exceptionally well, and it uses AI to handle everything else. Think of it as a creative partnership with AI as the production layer. The humans bring the vision, strategy, and taste. The AI brings the execution speed and breadth.

The second model is the collective. A loose network of independent practitioners who come together for specific projects and then disperse. Each member is a solo operator with their own AI toolchain and their own area of strength. The collective assembles the right combination of talent for each brief, without the overhead of permanent staff, office space, or holding company margin. This model has existed in some form for years in the freelance world, but AI makes it dramatically more viable because each individual can now deliver the output of what used to require a small team.

The third model is the solo practitioner. One person with exceptional taste, deep strategic thinking, and mastery of AI tools. This person can deliver — from strategy to execution to deployment — what used to require an agency team. Not for every type of brief. Not for every client. But for a growing and significant portion of the creative services market, the solo practitioner is not just viable. They are preferable.

I say this from experience. I have built eighteen products as a solo practitioner with AI tools. Each product involved strategic thinking, design decisions, content creation, technical implementation, and deployment. Each one would have been a standalone agency brief. None of them required a team.

What agencies should do

I am not writing this as a victory lap for the death of an industry I spent my career in. I am writing it because I believe the agencies that adapt will survive and thrive, while the ones that pretend nothing has changed will not.

The agencies that survive will do three things. First, they will radically reduce headcount and dramatically increase talent density. Instead of fifty people of mixed ability, they will employ ten people of exceptional ability, each amplified by AI tools. The team gets smaller but the output per person increases by an order of magnitude.

Second, they will stop selling hours and start selling outcomes. The billable hour is a relic of a model where time was a reasonable proxy for value. When AI compresses a week of work into an afternoon, billing by the hour becomes absurd. The new model bills for the value created, not the time spent. This is harder to price but more honest and more sustainable.

Third, they will invest in taste. The agencies that survive will be the ones that are definitively, obviously, undeniably better at knowing what is good. Not marginally better. Dramatically better. Because when the execution is democratised, the only remaining differentiation is judgment. The agency of the future is not a production house. It is a taste house.

The opportunity in the wreckage

Every disruption creates losers and winners. The losers in this shift are the large, slow, overhead-heavy organisations that sell scale and coordination. The winners are the fast, taste-led, AI-augmented operators who sell vision and judgment.

For anyone who has spent years inside agencies building skills in strategy, creative direction, brand thinking, or editorial judgment, this is the most exciting moment in a generation. The skills you have been accumulating — the judgment, the taste, the ability to look at a brief and know immediately what the answer should be — are about to become enormously more valuable. Not because the industry values them more, but because AI has removed every barrier between having those skills and deploying them directly.

You no longer need the agency to access the production capability. You no longer need the team to deliver the work. You no longer need the overhead, the hierarchy, the meetings, the politics. You need your taste. You need your strategy. You need an AI tool. And you need the courage to go.

The agency model is dying. But the skills it taught a generation of creative professionals are more alive than ever.

The question is not whether the model changes. The question is whether you change with it.