How we build
So much of health tech leads with the deck — confident slides, statistics without provenance, and a pilot booked before there's a working product to put in it. We took the harder order: build it, prove it on data, then partner.
Most
Chase a signed pilot to fund a product that doesn't exist yet.
Most
Big numbers on a slide, no method, no way to check them.
Us
A working prototype, a falsifiable hypothesis, figures that carry their provenance.
The falsifiable hypothesis
Claim. A hospital's largest recoverable losses are not clinical — they are coordination failures hiding in the seams between systems that never compare notes. Cross-index those streams and the losses become measurable, located to a unit, and recoverable.
Kill condition. If, on a partner's own data, the fabric cannot surface recoverable savings that clearly exceed our fee within a defined window, the hypothesis is false for that hospital — and we will say so, in writing, rather than move the goalposts.
A claim you can't disprove isn't a strategy; it's marketing. Ours has a kill switch on purpose.
The digital pilot
A read-only shadow deployment against a real hospital's historical streams — not a live experiment on patients, and not a survey. Every figure below is reproducible on your own data; that reproducibility is the point.
How we counted
Setting. One mid-size facility's historical data, cross-indexed by nOS. Read-only. No change to clinical operations during the window.
Method. Recoverable figures are tied to located stalls with a named cause — not modeled top-down from benchmarks. Conservative by default.
Honesty. Ranges, not false precision. Every number traces to a source and a confidence band, exactly as it does on the live surface.
Figures are stated as ranges because the honest version is a range. We'll walk your team through the method line by line.
For early collaborating partners
We're choosing a small number of early partners to build alongside. For them, pricing is results-based: our fee is a share of verified savings the fabric surfaces and you act on. If it doesn't find the money, you don't pay for the search.
It aligns the only way that's fair — we earn when you save. It also keeps us honest about that kill condition.
We cross-index your historical streams, read-only, and show what surfaces.
Your team validates the located savings against your own ledger. No black box.
Go live. Our fee is a share of what's verified and recovered — nothing more.
The constitution
These aren't aspirations on a wall. They're constraints in the code — some of them cost us features, speed, or a cleaner demo. We kept them because they're the difference between a tool clinicians trust and one they route around.
Every decision is judged by what it does to the person in the bed and the people caring for them. The patient is never "the bed" or "the case."
Resilience is not waste. A choice that drops a unit below its safety floor doesn't get a soft warning — it refuses, and shows its evidence.
No number without provenance. Predictions carry a confidence band. The system never pretends to know more than it does.
The safe choice is the easy choice. When the system is unsure, it fails toward caution — and tells you why.
The software absorbs the paperwork so the clinician can do the care. If a feature adds clicks at the bedside, it's wrong.
The system should get safer under stress, not more brittle. Downtime, surge, and the unexpected are designed for, not patched around.
Meet the hospital where it is. Vendor-neutral, additive, no rip-and-replace. We earn our place underneath, quietly.
Protecting staff is protecting patients. The system carries gratitude, names people, and treats the team as the point — not the cost.
Early partnership
We're taking a small number of partners into the shadow program. If the seams in your building are costing more than you can see, let's measure it.
partners@nightingale.os