Your Agency Already Runs on Composable Architecture. You Just Don't Know It Yet.

There's a 6,000-word whitepaper making the rounds right now from a company called Humaie. It's called "Designing the Intelligent Enterprise," and it lays out a framework for how companies should reorganize around AI. This is not the type of AI that simply involves adding a chatbot to your website. The structural kind. The kind that replaces coordination layers, collapses management hierarchies, and turns the org chart into a capability map.

It's dense and well-researched. 

And if you run a B2B agency between $1M and $10M in revenue, it should scare the hell out of you. 

Not because it describes a future you can't compete in. Because it describes a present you're already living in, and most agency founders haven't noticed.

Let me explain.

The paper's core argument is something called "composable business architecture." Strip away the consulting speak, and here's what it means: you should stop organizing your company around departments (marketing, sales, operations, and finance) and start organizing around capabilities. What can your company actually do? Not who reports to whom. What outcomes can you produce?

They break capabilities into three tiers:

Tier 1: Differentiating.
The stuff that makes you, you. Hard to replicate. Gets better over time. Your moat. Never outsource this.

Tier 2: Enabling.
This tier is important, but it is acceptable to be "on par" with competitors. Think HR, finance, and project management. Good candidates for AI augmentation or partial outsourcing.

Tier 3: Commodity.
It is so standardized that anyone can deliver it interchangeably. Automate it, outsource it, or let agents run it. Investing in excellence here is a waste.

Now. If you're an agency founder reading those three tiers, something should be clicking. Your agency already runs on a version of this architecture, whether you've thought about it this way or not. The problem is that most of you have the tiers in the wrong order.

Here's what I see when I work with agencies at this stage. Almost every one of them has, without naming it, built a composable architecture. They have:

A differentiating capability that's usually one of three things: a specific type of strategic thinking the founder brings, a delivery methodology that produces better outcomes in a defined space, or deep domain expertise in an industry vertical. This is the thing clients actually buy, even if the SOW says something generic like "digital transformation" or "brand strategy."

A set of enabling capabilities that keep the business running: sales process, project management, financial operations, talent development, and client communication rhythms. These are necessary, but not the reason anyone hires you.

And a pile of commodity capabilities that have been absorbed into daily work without anyone questioning why: basic reporting, time tracking, invoice processing, scheduling, template-based deliverables, and status updates that were created because someone asked for them once in 2019.

So far, so good. You're already composable. Congratulations.

Here's the problem: most agencies invest their founder's time, their best people, and their tightest margins in Tier 2 and Tier 3. The enabling and commodity work eats up 60-70% of the capacity, while the differentiating capability, the actual reason clients pay premium rates, gets squeezed into whatever time is left after the operational machine has been fed.

You built the architecture. You just wired the investment the wrong way.

The Humaie paper describes what it calls the "capability marketplace." The idea is that as AI agents become more capable, companies won't just automate internal tasks. They'll source entire capability layers externally, delivered by specialist AI providers and consumed on demand.

This isn't just a theory. It's already happening in the world of agencies.

Think about what's changed in the past eighteen months. Project management is getting eaten by AI-native tools that don't just track tasks but anticipate blockers, reallocate resources, and generate status reports without a human touching them. Financial operations for a sub-$10M agency can be handled by a fractional CFO plus AI-augmented bookkeeping at a fraction of the cost of a full-time ops person. Client reporting, the thing that eats Monday mornings alive, can be auto-generated from the same data sources your team manually pulls from every week.

And that's just Tier 3. Tier 2 is next.

Sales process? AI agents are already doing prospecting, research, initial outreach sequencing, and meeting prep better than most junior BDRs. Not because the technology is magical, but because the work is well-defined, data-rich, and pattern-based. Those are exactly the conditions where agents excel.

Talent development? The mentorship and culture-building part is irreplaceable (more on that in a minute). But what about the logistics of onboarding, training material delivery, performance tracking, and even skills gap analysis? All sourceable.

The Humaie paper describes three models emerging in this space: Capability-as-a-Service (standardized capability domains delivered by specialist providers), agent marketplaces (where you procure specialist AI agents for specific needs), and shared capability pools (multiple companies co-accessing commodity capabilities). All three are already showing up in the agency ecosystem.

Think about what the fractional CFO model actually is. It's Capability-as-a-Service for financial operations. What a white-label development partner is. It's a sourced capability for delivery execution. What a freelance strategist on retainer is. It's a composable capability slot filled by a human instead of a full-time hire. Agencies have been doing composable sourcing for decades. The difference now is that agents can fill those slots faster, more cheaply, and at higher volume for well-defined work.

The agencies that figure this out first, that deliberately externalize or automate their Tier 2 and Tier 3 capabilities, free up something that cannot be manufactured: founder attention, senior talent bandwidth, and margin to reinvest in the thing that actually differentiates them.

The agencies that don't will keep spending 70% of their energy on work that is about to become a basic requirement. And they’ll be baffled as their margins keep shrinking, even while the workload explodes.

Here's where the Humaie paper gets interesting for agencies and where I think most "AI is going to replace agencies" takes completely misses the point.

The paper cites sociologist Allison Pugh's research on what she calls "connective labor." It's the work of actually seeing another person, understanding their context, reflecting that understanding back to them, and building the kind of trust that allows real change to happen. Mentorship. Relationship management. In a client meeting, you can sense when the CMO's real concern is not the campaign performance, but rather her fear of being fired and her need for ammunition for the board.

No agent does that. No amount of prompt engineering does that. That is the differentiating capability for B2B agencies, and it's also the capability that gets most starved of investment because it doesn't show up on a timesheet or a utilization report.

I call this the hidden pipeline problem. Every agency I work with has a pipeline they can see: the deals in their CRM, the proposals out, and the contracts pending. But the pipeline that actually drives the business is invisible. It's the trust built in a client meeting that turns into a referral eighteen months later. It's the founder's pattern recognition that spots an opportunity the client hasn't articulated yet. It's the relationship equity that means you get called first when a new budget opens up, not because you submitted the best RFP, but because the client knows you understand their business.

That hidden pipeline runs entirely on connective labor. And it's the first thing that dies when your senior people are buried in Tier 2 and Tier 3 work.

The Humaie paper makes this argument in the context of Fortune 500 companies redesigning their management layers. But it applies with even more force to agencies, because the entire value proposition of a B2B agency is that you bring judgment, context, and human connection to a client's problem that their internal team can't generate on their own. If they could, they wouldn't hire you.

This means the agency that wins in the next five years isn't the one with the best AI tools. It's the one that uses AI to ruthlessly compress everything that isn't the differentiating capability, so that the humans in the organization can spend almost all their time on the work that only humans can do.

Strategy. Relationships. Judgment calls. The hard conversations. The creative leaps that come from understanding a client's business deeply enough to see what they can't see.

That's the moat. And it's actually a moat that gets wider with AI, not narrower. Because the agencies that free up their senior people from commodity and enabling work can invest disproportionately in deepening client relationships, building proprietary methodologies, and accumulating the kind of institutional knowledge that compounds over the years.

The Humaie paper drops some numbers that agency founders need to sit with.

Traditional SaaS companies generate about $200-300K in revenue per employee. A strong one might hit $500K. Midjourney, the AI image-generation company, generated $4.7 million in revenue per employee in 2025, with 107 employees. And then there's the extreme case: Medvi, a telehealth business that hit $401 million in revenue with two full-time employees. (The founder outsourced the entire regulated medical stack to partners and kept ownership of exactly one thing: the customer intelligence layer.)

You don't need to be Medvi. But you do need to understand what these numbers mean for the competitive landscape.

Right now, a typical agency with $3M in revenue has 15-25 employees. A fully optimized, composable agency at $3M might need 6-8 people, with AI agents and external capability providers handling everything else. That's not a theoretical exercise. I'm watching it happen in real time.

The founder who figures out how to run a $3M agency with 8 people and 65% margins isn't just more profitable. They're playing a different game. They can price more aggressively on the deals they want, walk away from the ones they don't, invest in the relationships that matter, and build a business that doesn't break when one senior person leaves.

The paper calls this "agentic leverage." I call it the math that will separate the agencies that survive the next three years from those that will get squeezed out.

If you've read this far and you're running an agency in the $1-10M range, here's the practical application. Not the theory. The work.

Name your differentiating capability. Not your service offering. Not your positioning statement. The actual thing you do that clients can't easily find somewhere else. If your answer is "we do great work," that's not how you stand out. That's a claim for a commodity. Go deeper. What specific type of judgment do you bring? What do you understand about your clients' world that nobody else in the room does? What gets better every year you operate?

Audit your time. For one week, track where your senior people's hours actually go. I guarantee that less than 30% is spent on the differentiating capability. The rest consists of enabling and commodity work that you have never questioned because you believe, "That's just how agencies work." It's not how agencies have to work. It's how agencies have worked because the alternatives didn't exist before. They exist now.

Draw the sourcing line. For every capability that isn't Tier 1, ask the question: Should we still be doing this ourselves? Not "can AI do this?" but "should a human in our organization be spending time on this task when that time could be spent on the thing that actually makes us money?" Some of this is AI tooling. Some of it is fractional hires. Some of it is just stopping doing things that nobody asked for in the first place.

Protect the human work. This is the one that gets missed. The whole point of compressing Tier 2 and Tier 3 isn't to run leaner for its own sake. It's to create space for the connective, relationship-driven, judgment-heavy work that is your actual product. If you automate away the admin but then fill that time with more clients at the same margin, you've missed the point entirely. The goal is fewer clients, deeper relationships, better outcomes, and higher value.

Build the governance before you scale. The Humaie paper is right about this: you need guardrails before you need more agents. What decisions require human approval? What does "good enough" look like for agent-generated output? Where does a human review before something goes to a client? Get this right early. Getting it wrong costs relationships, and in an agency, relationships are the whole ballgame.

For agencies specifically, governance has a layer that enterprises don't deal with: client trust. Your clients hired humans. They hired your humans, specifically. The moment an agent produces something that feels off, misses context, or sounds like it was generated by a machine pretending to understand their brand, you don't just have a quality problem; you now have a trust problem. And trust, once cracked, doesn't repair on a timeline that works for agency economics.

This doesn't mean you hide the AI. It means you own the standard. "Big G, little g" in an agency context means the founder sets the quality bar and client communication principles (Big G), while individual team members have the freedom to use whatever tools they need to get there (little g). The output belongs to the agency. The judgment about what's good enough belongs to humans. Everything else is negotiable.

The business world is just now realizing that the org chart is a 2,000-year-old way of routing information that AI is about to make obsolete. It’s such an important discovery that they're writing 6,000-word whitepapers about it.

Agency founders have been living this reality for years. You already run lean. You already compose capabilities across internal teams, freelancers, and partners. You already know that nothing can replace the founder's judgment and client relationships.

The question isn't whether your agency is composable. It is. The question is whether you're investing in the right tier. Because the capability marketplace is coming, and it's going to make Tier 2 and Tier 3 work radically cheaper and easier to source externally. The agencies that double down on their differentiating capability, the human judgment and relationships that no agent can replicate, will compound their advantage. Those who keep pouring their best people into enabling and commodity work will find that "the way we've always done it" has become the most expensive option on the table.

The Humaie paper ends with a quote from Block's Jack Dorsey: "What does your company understand that is genuinely hard to understand, and is that understanding getting deeper every day?"

That's the question. For your agency, today.

Bart Mroz is the founder of Agency Focus, a fractional growth consultancy for founder-led B2B agencies at $1M-$10M. He's spent 20+ years building, scaling, and selling agencies and now helps founders figure out what their business is actually good at and how to invest accordingly.

Reference: "Designing the Intelligent Enterprise," Cameron Smith, Humaie, April 2026. research.humaie.com/intelligent-enterprise.html