A new category of software
Most software is reactive. It waits for a person to tell it what to do: click this button, export that report, move this record. Even exceptionally good software still makes the human responsible for coordinating the workflow.
AI changes the unit of work. Instead of assigning a task, you can assign an outcome: “Reconcile last quarter’s expenses and flag anything outside policy.” The system interprets the objective, gathers the information, executes the steps and returns a result.
That is not a new feature. It is a different category of software. The winners will build from the agent outward, rather than bolt agents onto a SaaS product whose architecture still assumes a person is driving.
Why SaaS pricing breaks
Seat pricing assumes one human per licence. When one agent can do the work of fifty seats, that relationship no longer holds. The commercial model has to be rebuilt around outcomes, not access.
A public SaaS company can't rationally swap $50 per seat for an outcome contract worth a tenth as much in headline revenue, even if the new product is better for the customer. The installed base becomes a constraint, not an advantage.
New companies have no installed base to protect. They can price against the cost of the work, the value of the result, or both. That freedom matters.
The three-layer stack
1. Foundation. Model labs and compute. Enormous capital requirements, brutal return math, and capabilities that are increasingly interchangeable. We don’t invest here.
2. Orchestration. Frameworks, memory, evaluation and runtimes. We invest selectively, and only when the moat does not depend on a model lab choosing not to build the same thing.
3. Application. Vertical solutions that own a workflow end-to-end: billing for plumbers, intake for med-spas, dispatch for HVAC. Boring on the slide, enormous in the wallet. This is where we focus.
The application layer is where customers pay for a result. It is also where workflow knowledge, distribution and accumulated operating data can compound into a durable business.
Where defensibility comes from
A model API is not a moat. A prompt is not a moat. The durable advantages emerge when a product becomes part of how a business actually operates.
- Workflow integration depth — embedded in the full process, not a single isolated task.
- Proprietary workflow data — knowledge of how the work gets done and which interventions produce the best results.
- Outcome accountability — owning a measurable business result, with reliability customers can trust.
- Distribution lock-in — reaching a specific customer through relationships and channels that are difficult to recreate.
The historical precedent
Mobile took ten years to produce its defining companies. Cloud took twelve. AI will move faster: the infrastructure is already available, distribution is digital, and the customer can experience the return immediately.
That does not mean every AI company is a good investment. It means the window for thoughtful, operator-led judgment is open now. By the time the category feels obvious, the best businesses will not be priced as overlooked opportunities.
Consensus is what gets priced. Conviction is what gets paid.
Frequently asked questions.
What is agentic AI?
Agentic AI is software that can pursue an objective, select and execute actions, and monitor the result with limited human intervention. It owns a workflow rather than simply responding to a prompt.
How is it different from generative AI?
Generative is the engine; agentic is the vehicle. Generative AI produces content. Agentic AI uses that capability, alongside tools and business logic, to get work done.
Why does the application layer win?
The application layer sits closest to the customer, the workflow and the outcome. Revenue, proprietary workflow data and deep integration can compound there even as underlying models become interchangeable.
What does Cinque Capital invest in?
We partner with management teams to acquire and grow established AI and technology companies — majority and significant minority stakes in the lower middle market.