The Mispriced Systems Builders
The Boring but Critical Infrastructure That Turns AI Capability Into Business Value
This essay reflects a perspective on how value may be created in the current AI cycle. It is intended for informational purposes only and does not constitute investment advice or an offer to invest in any fund or security.
The Conversation Has Moved, Even If the Headlines Haven’t
Most of the public conversation around AI is still anchored in capability.
How powerful are the models?
What can they generate?
How quickly are they improving?
That made sense when the technology itself was uncertain.
It makes less sense now.
Because inside most organizations, the question has already changed.
It is no longer: What can AI do?
It is: What can we actually use?
That difference is subtle, but it’s where the next layer of value is forming.
Where Things Start to Break
If you spend time inside teams actually trying to use AI, a pattern shows up quickly.
Something works.
A team deploys a tool. A workflow gets faster. A task becomes easier. There is a moment where it feels like the promise is real.
Then the scope expands.
More users get involved. The system touches more parts of the workflow. Edge cases appear. Outputs vary. Ownership becomes unclear.
And what looked simple starts to feel unstable.
Not because the model stopped working.
Because the system around it wasn’t ready.
A generated answer still has to be trusted.
A recommendation still has to be owned.
A workflow still has to hold together when something goes wrong.
That’s where things slow down.
Not at the level of capability, but at the level of use.
The Real Constraint
It’s tempting to assume that better models will solve this.
In some cases, they will help.
But many of the constraints being surfaced are not model problems.
They are system problems.
How does this fit into an existing workflow?
Who is accountable for the output?
What happens when the system is wrong?
How is risk managed?
How is value actually measured?
These questions don’t disappear with more capability.
They become more important.
Which means the constraint has shifted.
From what AI can do → to whether it can be used.
The Layer No One Talks About Clearly
Once you see this, a different layer of the market comes into focus.
Not the models.
Not even the applications.
But the systems that sit in between.
The ones that make AI usable.
These systems don’t show up cleanly. They don’t belong to a single category. They tend to emerge in pieces:
workflow orchestration
trust and validation layers
governance and compliance
integration into existing systems
the mechanics of actually delivering and monetizing AI
Individually, each piece can look narrow.
Together, they form something more important.
The infrastructure of adoption.
Why This Layer Is Easy to Miss
There’s a reason this doesn’t dominate the narrative.
It doesn’t look like innovation in the way people expect.
It looks like:
process
coordination
integration
constraint management
In other words, it looks operational.
And early on, it often looks messy.
A tool solving a very specific problem. A system that only works in one context. A product that requires explanation.
It doesn’t feel like a category yet.
That’s exactly the point.
The Pattern Behind It
This isn’t unique to AI.
In most technology cycles, the systems that end up mattering most don’t look important at the beginning.
They show up where something breaks.
A workflow doesn’t scale.
A process becomes unreliable.
A constraint blocks further adoption.
At first, these look like local problems.
Over time, they reveal themselves as structural ones.
And the systems that solve them move from optional to necessary.
What “Systems Builders” Are Actually Doing
Some founders are drawn to these problems.
Not because they are obvious, but because they are unavoidable once you’re close enough to the work.
They don’t start with a category.
They start with a breakdown.
Something that doesn’t quite work, even though the technology exists.
And instead of building a feature around it, they start building the system that makes it function.
From the outside, that can look narrow.
From the inside, it’s usually where the real work is.
Why This Matters Now
We are at a point where AI capability is no longer the limiting factor in many contexts.
Which means the bottleneck is shifting into the system.
And systems don’t fix themselves.
They have to be designed, built, and refined.
That work is less visible.
But it’s what determines whether the technology actually gets used.
A Different Way to Look at the Market
If you focus only on capability, you end up asking:
What can AI do next?
If you focus on adoption, you start asking:
Where does AI break when it meets reality?
What has to exist for this to actually work?
Who is building that layer?
Those questions lead you somewhere different.
Every technology wave creates a moment where possibility outpaces reality.
AI has already reached that moment.
The models are here. The access is real. The experimentation is happening.
What comes next is not just more capability.
It is the work required to make that capability usable.
That work is quieter.
But it’s where things either come together or fall apart.
The next layer of value in AI may not come from what the models can do, but from what makes them usable.
If you’re building in this layer, or seeing similar patterns inside your organization, I’d be interested to hear how it’s showing up for you.
I write more about these ideas in Beyond Ventures™.
Compliance Note
This essay is provided for informational purposes only. It does not constitute an offer to sell or a solicitation of an offer to buy any securities. Any such offer will be made only through official offering documents and in accordance with applicable securities laws.


the perfect framing, would love to write more about this layer too. it's exactly the gap we keep seeing in our briefings
I'd been sitting with a version of this question for a while. You put it somewhere more useful.