AlayaCare

A home care management platform with intelligence layered onto it. The company's own description of the artificial intelligence is that it deploys directly into existing workflows and can sit over a customer's current technology stack, which is a statement about being an overlay rather than the substance.

Strip out the models and a complete business remains: care planning, scheduling, electronic visit verification, clinical documentation, billing, family portals, telehealth and mobile applications. That is what the company sold for a decade before the artificial intelligence products arrived, and it is what the majority of the roughly 608 person organisation builds.

The intelligence is nonetheless real and expanding, with a dedicated research group, two named products launched in 2025, and predictive risk modelling that carries a genuine clinical judgement rather than a workflow shortcut.

Graded C on the same reasoning applied to TeleTracking and Alcidion: healthcare only software, neither horizontal nor services, with authentic but non central intelligence. Watch the trajectory, since agentic workflow automation is where the company is investing.

Two very different autonomy postures sit in one product and neither is quantified.

The predictive side is advisory by design. Risk models flag clients before issues escalate and a clinician decides what to do, which is the appropriate posture for a hospitalisation risk prediction.

The automation side goes further. The company describes agentic workflows and round the clock agents that automate manual work and deliver operational savings, and claims agencies reclaim 80 percent of time and costs. Automating back office work at that scale means agents completing tasks without a human, and nothing published establishes which tasks, what proportion complete untouched, what happens on failure, or what a supervisor sees.

No accuracy figure appears for the risk models, which is the more consequential gap. A model predicting deterioration in home based clients has a false negative cost measured in avoidable hospitalisations and a false positive cost measured in wasted visits by a workforce already stretched, and neither rate is published. Ask for model sensitivity and specificity, and for the autonomy boundary on the agentic workflows.

Capability areas and product names are published, mechanism and measurement are not.

Named applications are specific: predictive risk of client deterioration, clinical documentation extraction, route and travel time optimisation, an employee retention model, and agentic back office workflows. Two products carry them, an assistant launched in Canada in June 2025 and an agentic workflow product launched in October 2025. A dedicated research group combining scientists and industry specialists is described as the source. The platform also ingests wearable device data for machine learning processing, which is an input disclosure most competitors omit.

What is absent is everything technical. No model class, architecture, foundation model, version or training data description for any component, and no accuracy, sensitivity or error figure for any of them.

The retention model deserves separate mention because it predicts on employees rather than patients, which is a different data subject with different expectations, and nothing describes what it uses or how it is validated. Graded C for named applications with no specification or measurement behind them.

The infrastructure layer is named and the model layer is not. The platform runs on a named public cloud, established through that provider's own published case material rather than through vendor marketing, and data sovereignty across the three markets is cited as a design consideration in the same source.

No foundation model provider, model class or version is identified for the assistant or the agentic workflows, both of which are very unlikely to be built from scratch. No sub processor list was located, and no statement addresses whether customer data, clinical documentation or wearable device signals contribute to model development.

The last omission is sharper than usual because of the range of data involved. The platform holds clinical documentation, visit verification records, wearable device streams and employee performance data across three jurisdictions with different rules, and whether any of it trains shared models is unaddressed in either direction. Competitors in this index answer that question explicitly and this one, which is otherwise strong on trust posture, does not. Ask for the model provider, the sub processor register, and the training position on customer and employee data.

Real deployments and no measurement of the intelligence.

Named customers include CBI Home Health, which renewed a national partnership in June 2025, and a cohort of 33 community support agencies modernised in August 2025. Operating scale spans three countries with roughly 608 staff, and a 50 million dollar growth facility from a Canadian bank in February 2026 indicates a lender satisfied with the underlying business. Aggregated user reviews cluster around 4.1 out of 5, with larger agencies rating the platform as worth the investment and smaller ones finding it heavy.

None of that measures the artificial intelligence. No published evaluation of the predictive risk models exists, no hospitalisation reduction study, no accuracy benchmark, no peer reviewed publication and no third party research organisation assessment specific to the intelligent components.

The headline claim needs particular caution. A stated 80 percent reclamation of time and costs through the newer automation is an extraordinary figure published in a March 2026 announcement with no methodology, no baseline definition and no customer attached. It should be treated as marketing until a source is produced. Ask which customers achieved it and against what baseline.

Better than most and short of the leaders, with the gap in a predictable place.

What is published: a trust centre as a standing destination, regular controls audits at the more demanding report type with the original certification conducted by a named international accounting firm, four privacy frameworks met across two jurisdictions, a stated commitment to transparency, security and responsible innovation attached specifically to the artificial intelligence, and reported customer controlled residency for protected information.

What is absent is the same pair that separates this band from the top. No retention schedule with stated periods, and no position on whether customer data contributes to model development or improvement. The second is conspicuous because the company is otherwise unusually forthcoming about governance, and because the data at issue is broad: clinical documentation, visit records, wearable streams and employee performance data.

The employee dimension deserves its own question. A retention prediction model operates on staff rather than clients, and nothing published describes what it uses, how long it is kept, or whether workers are told. Graded B for a genuine and externally assessed trust posture with the two central stewardship questions unanswered.

Four privacy frameworks are named individually across two countries, which is more explicit than nearly anything else in this index. The company states its artificial intelligence sits on a foundation meeting the United States health privacy and health technology statutes and the Canadian federal and Ontario provincial health information regimes, and that the platform passes regular controls audits at the more demanding report type.

Naming the two Canadian frameworks separately matters, because they impose different obligations from the United States regime on custodianship and disclosure, and most vendors operating in both simply cite the American one. A trust centre exists as a named destination for the underlying detail.

One capability reported by a third party directory, and worth confirming with the vendor, is customer controlled data residency for protected information, giving the agency control over where client data is stored for local regulatory compliance. If accurate that is a substantive privacy control rather than a policy statement.

Held at B because no business associate agreement posture, template or execution requirement was located, and because Australian privacy law is not named despite the company operating there. Ask for the agreement position by market and how the Australian regime is covered.

A real credential with an unusual amount of provenance attached, and a trust centre to sit it in.

The company obtained its first controls certification in 2020, and the announcement named the auditing firm, one of the large international accounting practices. Naming the assessor is rare in this index and is a meaningful disclosure, since it lets a buyer weigh the rigour of the audit rather than just the existence of a badge. Current material states the platform passes regular audits at the second report type, which covers operating effectiveness over a period rather than design at a point in time.

A trust centre exists as a named destination and is linked from the artificial intelligence material, and four compliance frameworks are named individually.

Held at B rather than A because the current position is stated without specifics. No audit period, current report date or availability process was located, the six year old announcement covers the earlier and weaker report type, and no penetration testing disclosure, vulnerability disclosure policy or health specific framework certification was found. Ask for the current report type, period and request process.

No device pathway is claimed and most of the platform plainly needs none. Scheduling, visit verification, billing, documentation and route optimisation are administrative.

One component sits closer to the line than the company acknowledges. Predictive analytics that flag clients at risk of deterioration, with the stated purpose of reducing hospitalisations, is software analysing patient data to support a clinical judgement, and that is the category where device definitions begin to bite. The exemptions in most regimes turn on whether the clinician can independently review the basis of the recommendation, which is design detail this record could not establish.

The question applies three times over. The company operates under United States, Canadian and Australian regulators, each with its own treatment of clinical decision support software, and publishes no position under any of them.

Graded C because the classification is probably correct and entirely undocumented, consistent with the treatment of Alcidion, the other multi jurisdiction record in this index. Ask for the device status determination in each market and the basis for any exemption relied on.

The first verifiable artificial intelligence governance credential recorded anywhere in this index.

The company publicly references its signatory status under a national voluntary code of conduct for advanced generative artificial intelligence systems, and links it from the same page that describes its artificial intelligence products. That code commits signatories to accountability, safety, fairness and equity, transparency, human oversight and monitoring, and validity and robustness. Signatory status is publicly listed and therefore checkable, which distinguishes it from the responsible artificial intelligence language other vendors assert with nothing behind it. A trust centre sits alongside it.

Held at B rather than A because a commitment framework is not a result. No bias or fairness testing output, no model validation methodology, no performance breakdown by population, no drift monitoring and no external audit of any model was located.

Two populations here warrant testing that is not disclosed. Risk models predicting client deterioration operate across three countries with different populations and documentation practices, and nothing addresses cross jurisdiction validation. And the employee retention model predicts on workers, where fairness obligations are employment law rather than health law. Ask what the code of conduct commitments produced in practice.

No performance figure is published for any intelligent component, so there is no stated level against which a shortfall could be measured. No accuracy, sensitivity, specificity or error rate appears for the risk prediction, the documentation extraction, the route optimisation or the agentic workflows.

No service level agreement, warranty, indemnity or remediation commitment was located. No pilot, trial or benchmark offer was found, and third party directories note the absence of any free trial or free tier, so a buyer has no published route to testing performance before committing.

The recourse question has a specific edge in this setting. A missed deterioration signal in home based care produces an avoidable hospitalisation or worse, and the client is not in a facility where staff might catch what the model missed. A vendor whose stated purpose for the risk models is reducing hospitalisations, publishing no sensitivity figure and no commitment, has taken on the claim without taking on any of the exposure.

One pre emptive note for future passes: further efficiency or savings claims cannot move this grade. The 80 percent time and cost figure measures operational benefit rather than model accuracy or recourse. Only a published performance measurement or a contractual commitment will change it.

Integration is described by category and never by name. The platform offers secure third party record system integration, ingests wearable and biometric device data for processing, and provides remote monitoring and telehealth alongside its own configurable clinical record, so it operates both as a system of record and as an integrator depending on deployment.

The artificial intelligence is explicitly positioned to work over a customer's existing technology stack without replacing it, which is an interoperability claim in its own right and implies the ability to read from systems the company does not control.

What is missing is every specific. No record system is named, no interface standard is described, no connection mechanism is specified, no device ecosystem is identified despite wearable ingestion being a stated capability, and no marketplace or certification listing was located.

Graded C for credible capability with no published specification, consistent with the treatment of Infinx, TeleTracking and Andor Health. Ask which record systems and device platforms are supported, through what standards, and whether integration differs by country.

One of very few records in this index where the residency question has an actual answer.

The platform runs on a named public cloud, established through that provider's own case material, which also cites data sovereignty as a design consideration supporting expansion across Canada, the United States and Australia. More substantively, a third party directory reports customer controlled residency for protected health information, giving the agency control over where client data is stored specifically to satisfy local storage regulations.

If confirmed, that is a materially better position than the silence recorded on this axis for most of this index, and it is the right architecture for a vendor operating across three regimes with different data localisation expectations, including Canadian provincial rules and Australian sovereignty requirements.

The sourcing caveat is real and the grade reflects it. The residency capability was located in a software directory rather than verified in the company's own material, and no region list, tenancy model or contractual residency commitment was found. Graded B on the strength of a named cloud plus a reported customer controlled residency capability, held below A pending vendor confirmation. Ask for the region list by market and whether residency is contractually committed.

No price is published by the vendor, and the economics are partially inferable from other sources.

The company publishes one directional economic claim, that agencies reclaim 80 percent of time and costs through its newer automation, which is the same shape as the cost reduction percentages that lift several competitors above D. It is unsourced and extraordinary, so it is credited as a disclosure and not as a fact.

Third party software directories consistently estimate entry cost from roughly 1,000 dollars per month, scaling with practitioner count, modules, integrations and training, with no free tier or trial. Those are estimates rather than vendor disclosure and are recorded as such, though the consistency of the scaling basis across independent directories makes the shape credible. One directory explicitly rates the company's pricing transparency as low.

What the vendor itself never states is the unit of charge, the module structure, whether the artificial intelligence products are included or priced separately, or how pricing differs across three countries with different funding models. The last matters most, since Australian disability scheme and Canadian public funding operate nothing like United States Medicaid. Ask for the licensing basis by market and whether the newer products carry incremental cost.

Enumerated with more precision than almost any record in this index, and genuinely multi jurisdictional.

The served segments are listed individually rather than gestured at: private duty nursing and non skilled care, Medicaid, community support services, therapy services, patient support programmes, clinical pharmacy services, hospice, home infusion, and Australian aged care including consumer directed care and the national disability insurance scheme. Naming a country specific funding programme is evidence of real localisation rather than a claim to be internationally available, because each of those regimes carries different assessment, documentation and billing rules that the platform has to encode.

Geographic coverage spans Canada, the United States and Australia, which places this among the small group of records here operating beyond one national system.

Held at B rather than A because scale by market is not disclosed, so the Australian and United States presence cannot be sized against the Canadian base, and because the reviews consistently indicate the platform suits medium to enterprise agencies rather than the whole range claimed. Ask for customer counts by country and by segment.