The New Middlemen

Welcome Back to XcessAI

For years, the dominant narrative around technology was disintermediation.

The internet would remove middlemen.
Platforms would connect producers directly to consumers.
Markets would become frictionless.

AI was expected to accelerate that process.

Instead, something else is happening.

AI doesn’t disintermediate.

It re-intermediates.

And the new middlemen may be more powerful than the old ones.

The myth of disintermediation

Every technological wave claims to eliminate intermediaries.

Search was supposed to remove brokers.
E-commerce was supposed to eliminate retailers.
Cloud was supposed to flatten infrastructure hierarchies.

What actually happened?

Value did not disappear.
It migrated.

New gatekeepers emerged.

Search engines replaced directories.
Marketplaces replaced physical retail.
Cloud providers replaced on-prem IT departments.

The interface changed.
The power structure did not vanish.

AI is repeating the pattern.

Why AI naturally creates middle layers

AI systems are powerful, but they are not self-contained.

They require:

• compute

• orchestration

• data pipelines

• integration layers

• workflow embedding

• distribution channels

That stack creates control points.

And control points become value pools.

The companies that sit between:

• users and models

• models and data

• models and execution

are becoming the new intermediaries.

These are:

• wrappers

• orchestration platforms

• agent marketplaces

• integration hubs

• API aggregators

• workflow controllers

They don’t build the core models.
They don’t own the end users.

But they control the interaction layer.

And interaction layers are where margins concentrate.

Value pools shift, they rarely disappear

In markets, value does not evaporate. It relocates.

When intelligence becomes abundant, scarcity moves elsewhere.

Scarcity now lives in:

• distribution

• proprietary data

• workflow lock-in

• trust

• embedded integration

As models commoditise, the differentiator is not raw intelligence.

It is control over:

• access

• routing

• sequencing

• visibility

The middle layer becomes strategic.

Not because it is technologically superior, but because it mediates power.

The quiet re-aggregation of control

There is a structural irony unfolding.

AI is marketed as decentralising.

But economically, it may concentrate power.

If every enterprise connects to models through:

• a small number of orchestration platforms

• a handful of API gateways

• a few dominant infrastructure providers

then intermediation increases, not decreases.

The surface area expands.
The dependency graph thickens.

New choke points emerge.

And whoever controls the choke points controls pricing, visibility, and leverage.

Why this matters for competition and margins

In competitive markets, margin durability depends on position in the value chain.

If you sit:

• upstream, building commoditising models

• downstream, competing on distribution alone

your margins compress.

If you sit in the middle, controlling access and integration, you gain leverage.

Middle layers benefit from:

• switching costs

• data aggregation

• multi-sided network effects

• coordination control

Historically, those positions have proven durable.

Think of payment processors.
Cloud platforms.
App stores.
Marketplaces.

They don’t eliminate friction.

They own it.

How middle layers capture margin

Intermediaries do not merely sit in the middle.
They shape how value flows through the system.

Control over routing creates leverage.

If you control which model is called, which vendor is prioritised, which dataset is surfaced, or which workflow is triggered, you influence:

• pricing power

• revenue share distribution

• visibility into transaction flows

• prioritisation of partners

• switching friction

Intermediaries can extract margin in subtle ways:

• through API pricing layers

• through revenue splits

• through preferred placement

• through data visibility advantages

• through sequencing control

When intelligence is abundant, routing becomes scarce.

And control over routing allows intermediaries to influence pricing, prioritisation, and ultimately margin distribution.

That is where the economic power sits.

The strategic question for enterprises

The real question is not:

It is:

And more importantly:

Every enterprise deploying AI must decide:

• Do we control our orchestration layer?

• Do we own the routing logic between models?

• Do we control our data access boundaries?

• Or are we renting intelligence through someone else’s gate?

That decision shapes long-term margin structure.

Why this phase feels different

Unlike prior tech cycles, AI intermediaries operate at the level of cognition.

They mediate:

• analysis

• recommendations

• prioritisation

• decision sequencing

That is a deeper layer of control than search ranking or retail shelf space.

The intermediary does not just influence distribution.

It influences judgement.

That amplifies its strategic weight.

Naming the shift

AI was expected to flatten markets. Instead, it is building a new coordination layer.

And coordination layers attract power.

Technology does not remove middlemen. It redesigns them.

The next dominant companies in AI may not be the smartest. They may be the ones who sit in the middle, quietly routing intelligence across the system.

That is where value is re-aggregating.

Until next time,
Stay adaptive. Stay strategic.
And keep exploring the frontier of AI.

Fabio Lopes
XcessAI

💡Next week: I’m breaking down one of the most misunderstood AI shifts happening right now. Stay tuned. Subscribe above.

Read our previous episodes online!