A client mentioned in a casual conversation last month that they had used an AI tool to do the audience research for their next campaign themselves. They told me this with the same tone someone might use to mention they had started cooking more at home — pleasant, slightly proud, no malice. They were not telling me to undercut my pricing. They were just sharing a thing they had done. The thing they had done was a piece of work I would have charged them several hundred dollars for six months ago.
I thought about that conversation for the rest of the week. Not because the client had done anything wrong, and not because I felt threatened by the specific instance. I thought about it because the conversation was a clear, casual indicator of something the agency industry has been quietly trying not to see. The bundle that agencies have sold for decades — strategy, production, distribution, accountability, relationship management, all wrapped together for a single retainer — is being disassembled in real time. The disassembly is not happening through dramatic disruption. It is happening through a thousand small moments like that conversation, in which a client or a freelancer or a small in-house team realizes they can do one of the things in the bundle themselves, with tools that did not exist eighteen months ago.
Almost nobody in the agency industry is talking about this honestly. The published conversation is still mostly about how AI will help agencies do their existing work faster, which misses the structural point entirely. The threat is not that AI will help agencies be more efficient. The threat is that AI will help clients, freelancers, and in-house teams do parts of the agency's work without needing the agency at all. This is a different problem, with different stakes, and the silence around it is itself the most telling signal that the industry has not yet figured out how to think about it.
This article is an attempt to think about it clearly.
Key Takeaways
- Agencies have historically sold a bundle. Strategy, production, distribution, accountability, and relationship management — combined into a single retainer that was easier to buy than the components separately.
- The bundle is being disassembled layer by layer. AI tools, specialist platforms, freelance marketplaces, and improving in-house capabilities are each unbundling specific parts of what agencies used to own.
- The middle of the agency market is most exposed. Boutique agencies with deep senior judgment and small single-operator studios with AI augmentation are both expanding. The mid-market generalist agency — competent but undifferentiated — is the structurally exposed segment.
- The published conversation is years behind the actual market shift. Agency industry coverage is mostly about AI as productivity gain. The harder conversation, about AI as disintermediation, is still mostly happening privately.
- The agencies that survive will not be the ones that bundle most. They will be the ones that own the layers AI cannot yet replicate — high-context strategic judgment, integrated decision-making, and long-term accountability for outcomes.
What The Bundle Used To Be
To understand what is being unbundled, it is worth being precise about what the bundle was. A traditional agency, in the model that dominated from roughly the 1990s through the early 2020s, sold clients a combined package of capabilities that were genuinely hard to assemble independently:
Strategy. Competitive research, audience analysis, content planning, positioning, brand voice development — work that required experienced people and significant time investment.
Production. Creative execution across formats — design, copywriting, video, photography, motion graphics — work that required specialized skills and expensive software.
Distribution. Channel selection, scheduling, paid media management, organic posting — work that required understanding of fragmented platforms and ongoing operational attention.
Accountability. Someone to call when results were not what was expected, someone to course-correct, someone to own the outcome rather than just the inputs.
Relationship management. A central point of contact who held the entire engagement together, prevented work from falling through cracks, and built the institutional knowledge that made each subsequent month easier than the last.
The bundle worked because the components were genuinely hard to acquire separately. Strategy required experienced strategists, who were hard to hire individually. Production required specialized talent across multiple disciplines. Distribution required tooling and ongoing attention. Accountability required a single throat to choke. Bundling all of this together at one agency was, for most clients, simply easier than the alternative of assembling and managing the components themselves.
This bundle is what is now coming apart.
How Each Layer Is Being Unbundled
The disassembly is not theoretical. It is observable, layer by layer, with specific tools and platforms doing specific work that used to require agency intermediation.
The Strategy Layer
AI tools now do a meaningful percentage of the strategic groundwork that used to require agency strategists. Audience research, competitive analysis, content gap identification, positioning exploration — all of these can be roughed out in an afternoon using ChatGPT, Claude, or Perplexity, at quality that was inconceivable two years ago. The output is not always polished, and it benefits significantly from experienced human direction, but it is no longer the case that producing a baseline strategic deck requires hiring a strategist.
What this means for agencies is that the strategy work which used to anchor the engagement — and justify the relationship — has had its commodity floor exposed. A client who could not have done their own audience research three years ago can now do it themselves in a few hours with a free or low-cost tool. The work that agencies do above that floor still has value, but the floor itself has moved up dramatically, and the agency has to demonstrate value above it rather than relying on the work below it to justify the engagement. [BACKLINK PLACEHOLDER → external: a credible industry analysis on AI in marketing strategy, e.g. from McKinsey, BCG, or a major industry publication's coverage of the AI shift. Aligns with the $5–12 CPC on AI tools and future of marketing.]
The Production Layer
The production layer is being unbundled across every creative discipline simultaneously. Midjourney, Stable Diffusion, and Adobe Firefly are doing image generation that used to require designers. Runway, Pika, and Sora-tier tools are starting to do video work that used to require editors. ElevenLabs is doing voiceover that used to require studios. Even copywriting, the discipline most resistant to AI for years, has had its commodity floor moved upward by GPT-4-class models and beyond.
The honest assessment is that the production layer is being unbundled fastest at the lower end and more slowly at the higher end. Mid-quality stock-style production is now effectively automated. Custom, high-craft work still requires human producers — but the line between "stock-style" and "custom" is moving steadily in the direction of automation, and the work that requires human producers in 2026 is a smaller category than it was in 2024.
The Distribution Layer
Native platform tools are absorbing the distribution work that schedulers and agencies used to provide. Meta Business Suite handles cross-posting between Instagram and Facebook with built-in analytics. TikTok's native tools have improved significantly. LinkedIn's analytics are now usable. The third-party scheduling and analytics tools that used to live between platforms and clients — Hootsuite, Buffer, Later, Sprout Social — are still useful, but the gap between native and third-party is narrower than it was, and clients can increasingly self-serve distribution without agency intermediation. [BACKLINK PLACEHOLDER → external: a comparison of native platform tools versus third-party schedulers, e.g. from Buffer or Hootsuite's own analysis of their evolving market.]
The Accountability Layer
This is the layer least disrupted by AI, and the one most agencies are not paying enough attention to as their durable competitive moat. AI cannot yet be on a call with a client when a campaign goes wrong. It cannot yet be the person whose reputation is on the line when results disappoint. It cannot yet maintain a relationship over years that holds together when individual projects struggle. The accountability layer is where the agency model retains its strongest structural value.
But this is also the layer most agencies are de-emphasizing in their own framing, because accountability is harder to market than capabilities. Agencies sell on what they can do, not on what they will be responsible for when something fails. That framing is exactly backwards in an AI-augmented future. The capability story is what AI is eating. The accountability story is what AI cannot yet replicate. Agencies that double down on the former are competing where they will lose. Agencies that build the latter into their core offer are positioning where they can win. [BACKLINK PLACEHOLDER → internal: link to article #7 (white-label invisible economy) — both pieces examine structural shifts in how the creative industry actually operates.]
The Relationship Layer
The relationship layer — the single point of contact, the institutional knowledge, the long-term partnership — is also relatively durable, but it is being squeezed by improved in-house teams. Companies that ten years ago needed an agency to manage their content presence increasingly have one or two in-house people who do most of the work, with agencies pulled in for specific spike capacity or specialized projects rather than ongoing retainer relationships. The relationship has not disappeared. It has been redistributed, with the long-term anchor moving in-house and the agency role becoming more project-based.
Who Survives This
The published agency conversation has mostly been about whether agencies survive. That is the wrong question. The right question is which agencies survive. The unbundling is not uniformly threatening. It is more threatening to some kinds of agencies than others.
The boutique end survives, and may expand. Small agencies built on deep senior judgment, narrow specialization, and integrated strategic thinking are exactly the layer AI cannot replicate. A boutique agency with five senior people who collectively own the strategy-to-production-to-accountability pipeline for a specific industry vertical is a more defensible business in 2026 than it was in 2020. The unbundling has not weakened this segment. It has clarified what makes it valuable.
The very low end also expands, in different form. Single-person studios augmented by AI tools can now produce mid-quality work that used to require small teams. These operators are not really agencies in the traditional sense — they are individual creators with significant leverage from tools. They are taking work that used to go to small generalist agencies, because they can deliver competent output at lower price points. The category is growing, even if the individual operators in it are not building durable businesses in the traditional agency sense.
The mid-market generalist is the exposed segment. Agencies with twenty to one hundred people, offering the full bundle to mid-sized clients at mid-market prices, are caught between the two expansions. The boutique end takes their high-value clients on quality. The AI-augmented single-operator end takes their lower-value clients on price. The middle has neither the depth to defend against the boutiques nor the cost structure to compete with the augmented solo operators. This is the segment where the unbundling is most visible in pipeline shrinkage and project loss, and it is the segment most often in denial about the structural nature of the threat. [BACKLINK PLACEHOLDER → internal: link to article #12 (agency margin trap) — both pieces examine the economic reality of mid-market agency operations.]
What This Means For Clients
If you are a brand or marketing leader trying to figure out how to structure your creative work in 2026, the strategic implications of the unbundling are different from what most agency conversations suggest.
First, the bundle is no longer the only option. You can credibly assemble strategy from AI tools, production from a freelance specialist or AI-augmented studio, distribution through native platform tools, and accountability through an in-house lead. This is more operational work than hiring a single agency, but it produces a meaningfully different cost structure and is increasingly viable for clients with even minimal internal capacity.
Second, the agencies worth hiring in 2026 are not the ones that do the most. They are the ones that do the things AI and freelance markets cannot do — senior strategic judgment, integrated decision-making across disciplines, and long-term accountability for outcomes. Paying agency rates for work that could be done with $50 of AI tools is no longer rational. Paying agency rates for work that genuinely requires senior human judgment still is.
Third, the agency conversation is still mostly happening in the framing of the old bundle. If you are evaluating agencies, the question worth asking is not "what can you do" — most agencies can do most things — but "what would I be hiring you specifically for that I could not get more cheaply elsewhere." The agencies that answer this question clearly are the ones worth working with. The agencies that respond with the full-bundle pitch are the ones that have not yet updated to the post-unbundling market.
