[ Our Thesis ]
On why the application layer endures — and which layers don't.
Every major industry runs on the same architecture: legacy systems of record that store current state, surrounded by human workforces that carry the actual decision-making logic in their heads. The architecture persists because of inertia, not because it works. Switching costs are high. Institutional knowledge is unwritten. The exceptions that drive real operations were never captured as data.
AI changes the equation. For the first time, software can sit in the execution path of complex workflows: not just recording what happened, but deciding what should happen next. When software delivers outcomes instead of tools, the addressable market expands from the software budget to the services budget.
But the harder question isn't whether AI disrupts legacy industries. It's which AI companies build durable advantages. Foundation models keep getting smarter. The application layer only endures where it has structural advantages the model layer cannot absorb.
A shift measured in trillions, not billions.
In healthcare, financial services, legal, and defense, there is a structural gap between what a foundation model can do and what a customer can legally accept. Someone has to own the compliance posture, the audit trail, and the domain-specific liability. Foundation model companies don't want that exposure. The platforms that absorb it become load-bearing infrastructure.
Legacy platforms stored what happened. AI-native platforms capture why. Every decision, every exception granted, every precedent applied generates a trace that makes the next decision smarter. Over time, those traces accumulate into a proprietary context graph — a living record of institutional reasoning no foundation model can replicate, because the model was never in the execution path.
Intelligence alone isn't a moat when the work is physical. Robotics, clinical devices, industrial automation, last-mile operations — these are domains where AI must be embodied, and where the integration of software, hardware, and real-world environments creates defensibility that can't be replicated from an API endpoint.
[ Where We Invest ]
The same advantages compound differently in each vertical.
Clinical decision-making, revenue cycle, care coordination.
Underwriting, compliance, the operating layer of capital.
Workflow systems for high-context, high-stakes decisions.
Coordinating fleets, inventory, the physical movement of goods.
Mission systems where accountability is the product.
The operating layer for properties, portfolios, tenants.
We invest at the intersection of these advantages. We look for AI-native platforms in industries defined by labor intensity, regulation, exception-driven decisions, and physical complexity. These companies don't automate what humans used to do. They build the institutional memory and regulatory standing that makes them irreplaceable: to incumbents, to competitors, and to the model providers themselves.
[ Intrigued? ]
We're always looking for exceptional founders building category-defining companies in our focus areas.
intrigued@inertia.vc