AlayaCare

The platform carries its own value, and one of its two named products runs on models. AlayaCare Cloud is a full home care operations and clinical record platform covering intake, scheduling, visit verification, care planning, documentation, billing, payroll and family portals, and it sold on that for a decade before its AI products arrived.

AlayaFlow, now presented beside AlayaCare Cloud as the second platform, is an agent layer that detects events such as failed or vacant visits and resolves them, drafts care plans from assessments and handles referrals; the company says it works over AlayaCare Cloud and other home care record systems. Models also drive the Client Intelligence Suite's risk prediction, the Layla assistant, voice-to-form dictation and smart summaries. The agents are a core module that does not exist without the models, while the platform around them would still sell.

A correctness promise sits beside a real review structure, and the promise undercuts it. AlayaCare's March 2026 announcement says the Recommended Care Plan Agent 'ensures all recommendations align with agency protocols and existing care libraries', and its security page offers agents that 'eliminate visit verification errors'. An agency that believes either will not build the review step that catches the miss, whatever else the product offers.

What it offers is substantial: agencies configure whether each agent drafts suggestions or executes automatically, every decision is logged with an explanation, the care plan agent leaves the clinician the final sign-off, and agents escalate to staff when they cannot resolve a case.

The AI page says every AI action is logged and human-approved, while the AlayaFlow page describes workflows that resolve without escalating to the team, and the company counts work resolved without manual effort (over 80 percent of failed visit verifications, over 70 percent of scheduling disruptions); both cannot hold for an agent set to execute. The company publishes neither the conditions that send a case to a person nor the error rate on work completed without one.

The modeling approach is named, and no model or version is. AlayaCare's AI page says it uses parameter-efficient fine-tuning and smaller task-specific models, keeps client data isolated for training, testing and inference, and uses a fairness-through-unawareness design that avoids bias-sensitive data.

Company material also describes large language models extracting clinical information from unstructured documents and summarizing change in condition, and AWS's published case study says the Layla assistant is built on Amazon Bedrock. Named applications include hospitalization and adverse event risk prediction, clinical note detection, visit and route optimization, voice dictation into forms, smart summaries and the AlayaFlow agents. No base model, version, update cadence or notice of model change is published, and no accuracy figure accompanies any component.

The host is named by AlayaCare and the model layer is not. The company's performance and security page says the platform runs across multiple AWS availability zones as AWS Partner Qualified Software, and the security page says each customer's records are hosted in its region. For the model layer, AlayaCare's own pages say only that it fine-tunes smaller task-specific models.

AWS's published case study says the Layla assistant is built on Amazon Bedrock, which would keep model processing inside AWS, but AlayaCare names no model provider or foundation model itself. The company publishes no subprocessor list. The data involved spans clinical documentation, care plans, visit verification records, caregiver calls taken by AI call operators and employee data across three countries.

Named customers and vendor figures, with no method behind them. Case studies credit Enliven Home Infusion Specialists with 200 percent revenue growth, 400 hours of monthly admin time saved and 40 percent faster billing processing, and First Day Home Care with 175 percent year-over-year revenue growth, 187 percent growth in billed hours and a 90 percent clean claims rate; Elara Caring, Option Care Infusion and Avid Health at Home appear among customers on the home page.

For the AlayaFlow agents the company cites over 80 percent of failed visit verifications resolved, over 70 percent of scheduling disruptions resolved without manual effort, and up to 80 percent savings in time and cost. The care plan agent's documentation saving appears three ways: up to 20 percent on the November 2025 webinar page, up to 50 percent in company announcements, and 50 percent or more in the March 2026 webinar deck. None of these figures carries a baseline, period, denominator, method or named deployment. Nothing is published evaluating the hospitalization and adverse event risk models.

A categorical training commitment is published without the retention schedule behind it. AlayaCare's AI page says client data is fully isolated for training, testing and inference, that AI features follow the platform's role-based permissions, and that model design avoids bias-sensitive data. The security page says records are hosted in the customer's region and encrypted in transit and at rest.

Not stated: whether a customer's data trains models used for other customers or only its own, what is kept from AI interactions and for how long, and whether data is de-identified before any shared use. Safety engineering, meaning guardrails, hallucination mitigation and how a safety event is handled, is not described, although the care plan agent and the Layla assistant both produce clinical content.

Status is stated for each market and supported by published assessments, while the agreement itself is not published. AlayaCare says its platform complies with HIPAA and HITECH in the United States, PIPEDA and provincial statutes in Canada, and the Australian Privacy Act, and that it undergoes annual HIPAA audits.

Its security page says it publishes Threat and Risk Assessments and Privacy Impact Assessments, which sit with the audit reports in its trust center, and describes consent directive and minimum-necessary controls agencies can apply to client records. Neither the agreement's terms nor the subcontractors it binds appear on the site.

AlayaCare names current attestations of its own, and the reports sit behind its trust center. The security page lists independent SOC 1 Type II, SOC 2 Type II, HITRUST i1 and annual HIPAA audits, and a February 2026 announcement says AlayaCare Cloud achieved HITRUST i1 certification.

The page says the SOC, HITRUST and HIPAA reports, security whitepapers and assessments live in the trust center and are available on demand, and the trust center itself may show whether they open without a request. The company's first SOC 2, a Type 1 in 2020, was performed by PwC. Encryption, multi-factor authentication, single sign-on and audit logging are described.

No regulatory position is stated in any market, and the one clinical output restates the agency's own library for a clinician to approve. Scheduling, visit verification, intake, billing and documentation are administrative, and the Client Intelligence Suite's ranking of clients by predicted hospitalization and adverse event risk is population health work.

The Recommended Care Plan Agent drafts diagnoses, goals and interventions: the launch release says it works from caregiver input and client data, and the March 2026 webinar deck says it interprets assessments, diagnoses and medical history already in the chart. The company says its recommendations are built on the agency's own care plan libraries, which the agency writes, and a clinician gives the final approval, so the output restates configured content rather than composing care.

The release also says the agent generates personalized goals and interventions, and nothing says whether it can propose anything outside the library. AlayaCare operates under US, Canadian and Australian regulators and states a position under none of them.

A public governance commitment with a described bias approach, and no results. AlayaCare's AI page says it is a signatory of Canada's Voluntary Code of Conduct on the Responsible Development and Management of Advanced Generative AI Systems, whose signatories commit to measures on accountability, safety, fairness, transparency, human oversight and robustness, and that every AI action is logged.

It describes its bias approach as avoiding bias-sensitive data, using a fairness-through-unawareness model design and running ongoing fairness checks. That is a framework with a named process behind it. Nothing is published from those checks: no results by client group, service line, language or country for the risk models or the care plan agent. Fairness through unawareness can leave bias in place where other inputs stand in for the omitted ones, which matters most for the hospitalization risk models.

Mechanisms let an agency challenge an agent's work, and nothing stands behind work that is wrong. AlayaFlow logs every agent decision with an explanation, offers observability and traceability across actions, can be set to draft rather than execute, and leaves the clinician the final sign-off on care plans.

The March 2026 care plan agent webinar deck prints 'Outcome Based Pricing' beneath the three agents and their headline figures (over 80 percent of failed visit verifications resolved, 50 percent or more less documentation time, over 70 percent of scheduling disruptions resolved without manual effort), but defines no outcome, fee, baseline or share of the fee at risk, so what an agency pays when an agent misses is not stated.

No warranty, indemnity, service credit tied to output quality or measured error rate is published. The stakes differ by agent: an unfilled visit leaves a client without care, a wrongly resolved visit verification exposes the agency on Medicaid billing, and a care plan suggestion the clinician accepts becomes the plan.

AlayaCare is itself a full home care record system, and it describes its connections to other systems without naming them. The platform holds clinical documentation, care plans, medications and visit records, and that record is the product itself rather than an integration. The company describes a Connector, API management, a marketplace of partner extensions and data movement between systems, and says the AlayaFlow agents work across AlayaCare Cloud and other home care record systems.

A customer on the home page credits AlayaCare with Waystar for claim submission, and a company news release describes a planned hospital-to-home integration with St. Joseph's Healthcare Hamilton. None of these names a record system, health information exchange or interface standard with a production deployment behind it.

Residency is stated as a commitment and the processing path is left open. AlayaCare's security page says every client record is hosted in the customer's region, and its performance page describes AWS infrastructure across multiple availability zones with automatic failover and segregated, encrypted backups. The AlayaFlow page states tenant isolation for the agents. The site gives no region list by market and does not say where AI processing happens.

AWS's case study says Layla is built on Amazon Bedrock with data residency controls, but nothing from AlayaCare says whether model calls stay in the customer's region, which matters most for Canadian and Australian customers whose rules on cross-border transfer differ from the US.

No price is published, and AlayaCare's own pricing page sets out how the product is sold. Pricing is custom, based on the organization's size, care models, regions and the modules chosen. Three packages (Building and Upgrading, Growing and Scaling, Transforming and Optimizing) are shown with the modules each includes, among them the intake, scheduling, billing and clinical AI agents.

A base subscription typically covers scheduling, visit management, documentation, workforce tools and standard reporting, caregiver and supervisor mobile apps are included in most configurations, and there is no free trial. The company's March 2026 agent webinars label the AlayaFlow agents 'Outcome Based Pricing' without saying what counts as an outcome or what a customer pays when an agent fails. No figure, unit of charge or implementation fee is published.

Coverage is named service line by service line and market by market, with validation behind part of it. AlayaCare sells in the United States to home care, home health, home infusion, private duty nursing, pediatric care, intellectual and developmental disability services, Medicaid and Veterans Affairs programs and long-term care insurance billing, and to home and community care in Canada and aged care in Australia and New Zealand, from single locations to enterprise networks.

Named customer stories sit behind home infusion (Enliven) and home care (First Day), with other agencies named on the home page; nothing sizes the customer base by country or service line. For intake and scheduling work, the Intake module manages referrals and intake, the AlayaFlow Intake Agent receives, screens and scores incoming referrals against caregiver availability, and a Scheduling Agent fills vacant visits.