Still, 16:9: the front of a small independent agency office, morning, a sign in the window, a woman unlocking the door
Bootstrapped. 130 Medicaid home and community-based care agencies. A 99% clean-claim guarantee, in writing. AI infrastructure priced on care revenue, with margins that expand as exceptions fall.
Leave your details and we send two things by email within one business day: a recorded briefing by Anand Chaturvedi, and the investor brief as a PDF, with the operating figures behind the six measures on this page, each with definition and source.
The recorded briefing and the PDF brief will reach your inbox within one business day. If you would like the operating figures, reply to that email and we will send the NDA.
Sources: CMS/Mathematica LTSS expenditure brief (2023); PHI Direct Care Workers Key Facts 2025; Caryfy estimates for agency-billed and segment shares, to be anchored by the CareDrain methodology paper.
By mandate, by rate cut, by demographics. All 50 states reported home and community-based care workforce shortages in 2025. An agency inside a compressing envelope cannot carry a human administrative layer and cannot pay a vendor indifferent to outcomes. No single rule and no single date is load-bearing; the direction is.
Cary®, the orchestrating agent, and nine Care Apps on one ledger. The work leaves the owner's hands; authority stays human. Confinement architecture, confirm-back before every write, append-only records, and CareLore turning every resolved exception into a reused rule.
The architecture in detailWhat flows through the ledger; grows with agencies added and tiers stepped up.
Blended share of eligible billed revenue across the base, with the platform plan beneath it.
The share of work that still reaches a person. Falls as CareLore turns resolutions into rules.
Expands as the exception rate falls; the human layer is a cost of delivery, not a revenue line.
Agencies step up tiers as they hand over more of the work.
The impact metric: caregiver hours moved from screens to patients.
The humans are not the scale unit. The infrastructure is. The human layer is a cost of delivery; the share of work that reaches a person falls as the network learns.
Current values, definitions and sources are in the investor brief, shared under NDA.
Every agency starts on the full platform. Each step up is a decision to stop operating a function and start receiving it. The trigger is the same every time: the owner sees the work carried in one function and asks for the next.
Frontier models are trained on the internet. Home and community-based care is not on the internet. It is a domain where every action gets a verdict — the payer accepts the claim or rejects it, the visit verifies or it does not, the survey passes or fails — and where almost no model has seen the data. High verifiability, low training attention. That is the combination that rewards a system which learns against ground truth on its own operating base.
Payers, EVV systems and surveyors return a verdict on nearly every action. The system is evaluated by the market, not by a benchmark.
Program-by-program rules, denial patterns, authority boundaries and resolved exceptions, captured with outcomes across 130 agencies and growing.
The model reasons inside fixed phases; authority runs as tools. We know exactly what should be autonomous — and exactly what should never be. That is why a regulator can audit it and why a guarantee can sit on top of it.
Resolutions captured and reused. The exception curve is the proof. New entrants start at the top of the curve.
Confinement architecture and audit-grade trails. Regulators and payers can inspect what the system did and who signed. A wrapper cannot retrofit this.
AI carries most of the customer journey, from first diagnostic to onboarding. Acquisition cost does not scale with customers.
Enterprise AI platforms reach production by deploying engineers on site, then handing over. That model cannot serve 28,000 single-location independents; the deployment cost exceeds the account. The only way into this market is a system that carries its own onboarding — which is what the distribution moat is.
Vertical operating systems that take a share of the flow they carry — restaurants, home services, payments — proved that pricing on flow earns a software multiple when the cost base is software. Hospital revenue-cycle outsourcing proved the percent-of-revenue fee logic for one function at the top of the market, with a services cost base and a services multiple. Autonomous Care OS® applies flow pricing to nine functions for 28,000 operators no services firm could serve profitably — and puts the fee at risk: a 99% clean-claim number guaranteed in writing, with automatic fee credits when missed. A cost-plus vendor cannot copy that without becoming us.
Our only target segment through 2028. Growth from tier upsell in the base and geographic expansion across Medicaid states.
Price discovery and category validation for Europe. Private Spitex operators, three payers per client, no mandate. Self-funding by design; the first stop on a global roadmap.
Operating footprint today: United States. Switzerland enters the footprint when revenue makes it true.
Home of CareBravo®, in operation across 130 Medicaid home and community-based care agencies. Operations, billing and customer delivery; where the ground truth is generated and where the guarantee is honoured.
1360 Peachtree St NE, Suite 800, Atlanta, GA 30309Sovereign, multilingual care AI built on Apertus with Swiss institutions; private Spitex operators as the test population. Platform and engineering are built in India; clinical content is reviewed by a named clinical reviewer before it ships.
Corniche Road 1, Metio Building, 1066 Epalinges, VaudTwo decades in long-term care technology. Built the SaaS, watched it relocate the burden, rebuilt from first principles. Author of The Care Manifesto and architect of Work as Services and Autonomous Caring®.
anandchaturvedi.comThrough CareBravo: a platform plan per active client and percentage plans up to 4.47% of eligible billed revenue. Revenue follows care delivered; cost follows exceptions; exceptions fall as the system learns.
About 28,000 US Medicaid home and community-based services agencies. US Medicaid HCBS spending was $145.9 billion in 2023, up 50% since 2019. Agency-billed all-payer HCBS revenue is estimated at $140–170 billion, of which Caryfy's target segment bills an estimated $45–55 billion.
Data: every resolved exception is captured and reused across the network. Authority: a confinement architecture whose audit trail regulators and payers can inspect. Distribution: AI carries most of the customer journey from first diagnostic to onboarding.
Caryfy Inc. in Atlanta operates CareBravo in the United States. Caryfy Sàrl in Lausanne, Switzerland conducts price discovery and category validation for Europe and research on sovereign, multilingual care AI.
Request the investor brief on this page or at investors@caryfy.ai. Operating figures are shared under NDA.