The Thesis
The architecture HCBS has been waiting for.
Why operations cannot be fixed by better software — and what the model+services pattern, validated at horizontal scale in May 2026, means for the care economy.
I.
Home- and community-based services agencies have been buying software for twenty years.
Scheduling systems. EVV platforms. Electronic medical records. Billing systems. Compliance trackers. Caregiver mobile apps. Family portals. Payroll integrations. Each tool, individually, was supposed to solve a problem. Each tool, individually, did solve some version of its problem. And yet, by every measurable indicator, the operating reality of running an HCBS agency in 2026 is harder than it was in 2016.
Margins compressed to 5–8%. Caregiver turnover at 77–80%, with seven in ten new hires gone within a hundred days. Agencies using three or more disconnected tools lose 14 hours per week to manual reconciliation. CMS finalized a 1.3% reduction in aggregate Medicare home health payments for CY2026. HR 1 and the One Big Beautiful Bill Act have introduced Medicaid eligibility and reporting pressures with state-by-state ripple effects that are still arriving.
The data is not ambiguous. Better software, by itself, did not fix the operating model. It is increasingly clear that it cannot.
II.
The reason is structural, not a failure of any particular vendor.
Software as a Service has a defining property: it sells the buyer a tool the buyer must operate. Every SaaS purchase, by design, is a request that the buyer take on more operational responsibility. The vendor delivers the tool. The operator staffs the tool. The operator runs the tool. The operator handles the exceptions. The operator absorbs the integration cost. The operator absorbs the training cost. The operator absorbs the cost when the tool doesn't actually do what was promised.
This model works in industries where the cost of operating the tool is small relative to the value it creates. It breaks in industries where operating the tool is the work — where the cost of running a stack of disconnected tools approaches or exceeds the value any individual tool creates.
HCBS is exactly that industry. The cost of operating the back office is now consuming the margin the operation is trying to produce.
III.
The architectural shift that solves this is not better software. It is a different category entirely.
Work as Services inverts the SaaS contract. Instead of selling the operator a tool, the substrate delivers the work itself. The software, the agents, the human expertise, and the institutional knowledge required to operate the back office are bundled into a single substrate that the operator does not run. The operator buys the outcome. The substrate operates.
This is not a managed-IT model. Managed IT keeps your servers running; the work of running your business is still yours. This is not a BPO model. BPO offshores tasks to lower-cost labor pools; the operating model is still yours to design. This is something architecturally newer: a substrate composed of software, autonomous agents, and dedicated humans, operating as one system, owned and accountable end-to-end, designed from first principles for one industry's operating reality.
The reason this category became possible in 2026 — and was structurally impossible in 2020 — is that the underlying intelligence required to run the agent fleet became real. Frontier-grade models, vertically fine-tuned (in our case on the open-source Apertus model from EPFL and ETH Zürich, through Care OS™'s Panacium intelligence layer), can now do operational work that previously required a person.
The dominant AI labs in the world have recognized this, and have built businesses around it.
The Validators
$5.5 billion. Eighteen days. One thesis.
In May 2026, the two leading frontier-model laboratories on the planet — and a consortium of the world's most influential private capital firms and consulting practices — collectively committed more than five billion dollars to a single, structural bet: the frontier model alone is not enough. The model must be embedded inside the operation, by humans accountable for the outcome, to actually deliver value at enterprise scale.
May 4, 2026 · San Francisco
Anthropic + Blackstone + Hellman & Friedman + Goldman Sachs
Anthropic, Blackstone, Hellman & Friedman, and Goldman Sachs announce the formation of a $1.5B AI-native enterprise services firm. The firm embeds Claude directly into the core business operations of mid-sized companies in the consortium's portfolios. Anthropic engineering and partnership resources are embedded into the firm's team. Backed additionally by General Atlantic, Leonard Green, Apollo Global Management, GIC, and Sequoia Capital.
The Fortune headline: "Anthropic takes shot at consulting industry."
May 11, 2026 · San Francisco
OpenAI Deployment Company
OpenAI launches the OpenAI Deployment Company. $4B in initial investment, $10B valuation, 19-firm investor consortium led by TPG and including Goldman Sachs, SoftBank, Bain Capital, Brookfield Asset Management, Warburg Pincus, BBVA, and B Capital. Named consulting partners: Bain & Company, McKinsey & Company, Capgemini. Acquires Tomoro, a 150-engineer applied AI firm.
The model: Forward Deployed Engineers (FDEs), borrowed directly from the Palantir playbook. Engineers parachute into client organizations and live inside the complexity — legacy infrastructure, compliance constraints, convoluted permissions — rather than shipping software and leaving the implementation to someone else.
Both ventures are explicitly horizontal. Their thesis: any industry, any function, any size of mid-market enterprise — wherever there is operational complexity that AI can absorb, embed humans and agents together inside the operation and deliver the outcome.
Careonomy is the vertical execution of the same thesis. Where Anthropic's enterprise services firm and OpenAI's Deployment Company go horizontal, Careonomy goes deep on one industry: home- and community-based services. Where Forward Deployed Engineers parachute into Fortune 500 IT environments, Care Business Advisors® embed alongside HCBS operators and the Care OS™ agent fleet that runs their operation. Where the horizontal firms validate the pattern, the vertical execution earns its right to exist by being purpose-built for one industry's regulatory, workforce, and operating reality.
IV.
The contrast between SaaS and Work as Services is more than commercial. It is architectural.
| Dimension | Software as a Service | Work as Services (Careonomy) |
|---|---|---|
| What the vendor sells | A tool to operate | The operation itself |
| Who operates the tool | The buyer | The substrate — software, agents, and Care Business Advisor® together |
| Who absorbs operational exceptions | The buyer's staff | The substrate's agent fleet and Care Business Advisor® |
| How cost scales with growth | Linear: more clients require more staff to operate the tool | Sub-linear: the substrate's marginal cost-to-serve approaches flat |
| What success is measured by | Tool adoption and feature usage | Operational outcomes: hours reclaimed, denial rates reduced, retention improved |
| Who is accountable for outcomes | Nobody contractually — the vendor is paid regardless of outcome | The substrate provider, through Outcome Assurance™ |
| The buyer's role over time | Operating their own back office, with better screens | Strategy, growth, payer relationships, clinical leadership — not operations |
V.
The Care OS™ architecture is what makes the vertical execution possible.
Anthropic's enterprise services firm has Claude. OpenAI's Deployment Company has GPT and the Frontier platform. Both have horizontal frontier models that are extraordinary at general reasoning, and both have to fine-tune those models, integrate them with legacy systems, and configure them per-customer for every vertical they enter.
Careonomy starts from a different place. The intelligence layer — Panacium, built on and forked from Apertus, the open-source Swiss large language model from EPFL and ETH Zürich — is purpose-fine-tuned for the regulatory, clinical, and operational reality of HCBS. The technology product layer — CareBravo (the registered Autonomous Care OS®) — is purpose-built for HCBS operations. The category and media layer — Karavista — is purpose-built for HCBS operators and the community that surrounds them.
Careonomy is the layer where these come together into a single operating partnership: the software that knows your industry, the agents that have been trained on your regulatory environment, the human expertise that has run agencies like yours, and the institutional learning that compounds across every operator on the substrate.
That's what a vertical Work-as-Services substrate looks like. And that's what Careonomy is.
VI.
The next move.
The 4A Framework™ is how Careonomy engagements begin. Twelve weeks. Four phases — Assess, Align, Activate, Accelerate. Delivered by a certified Care Business Advisor® alongside the Care OS™ agent fleet that will eventually operate your back office.
Or, if you want the briefer first pass: the Provider Blueprint™ is the 90-day strategic assessment and implementation plan that anchors the engagement.