Honor

AI is the scaling mechanism of the operations platform: the CEO states more than 20 distinct interoperating AI algorithms run in the network, doing caregiver to client matching and scheduling. Held back from A because the deliverable is a home care operations service in which AI optimizes matching rather than being the product a buyer purchases outright.

The system schedules and matches, and how much of that happens without a person is not stated.

Nothing located describes whether a human coordinator reviews or can override a proposed match, whether allocation runs automatically or as a recommendation to office staff, whether a care professional can decline an offered shift without that decision affecting future allocation, whether either party is shown why a match was made, or what recourse exists when the system gets it wrong.

For most records in this index the oversight question concerns clinical output. Here the system's output is a work assignment given to an employee, which makes the absence of a described override path and appeal route more consequential than usual. An automated allocation that a worker cannot see the basis of and cannot contest is a different proposition from a scheduling aid a coordinator confirms.

The questions to put are concrete. Is a match auto assigned or proposed. Who can override, and is the override logged. Does declining carry a consequence, stated or unstated, for future offers. And is there a person accountable for reviewing allocation patterns over time rather than individual matches.

What is public is a count, not a description. Company leadership has stated that more than twenty distinct interoperating algorithms run in the network, performing caregiver to client matching and scheduling. A count of algorithms is not a description of any of them.

Nothing was located describing what any model takes as input, what it produces, how matching is scored, how the system was trained or on what, how often it is updated, or how performance is measured. There is no technical documentation, no model card, no versioning statement and no product documentation available to a prospective agency partner.

The claim itself is also the kind this index flags. A stated quantity of algorithms is a scale signal that reads as sophistication while remaining unfalsifiable, because no reader can check a count of things none of which are described. Ask instead what the matching model optimises for, since that single answer tells an agency more than the number of models does.

Nothing identifies any party in the chain: no model or model family, no foundation model provider, no hosting arrangement and no sub processor list was located in two passes, and no technical documentation, model card or versioning statement is available to a prospective partner. What is public is a count rather than a description. Company leadership has stated that more than twenty distinct interoperating algorithms run in the network.

A stated quantity of algorithms is a scale signal that reads as sophistication while remaining unfalsifiable, because no reader can check a count of things none of which are described, and it belongs with the other claim shapes this index tracks that cannot be wrong in any way a reader could establish.

One supplier relationship is visible in passing and it is on the worker side: the care professional application uses a third party mapping interface for geolocation, disclosed by reference to that provider's own terms. That is a party in the chain identified indirectly, and it is the only one. Nothing states what location data is used for beyond dispatch, how long it is kept, or whether it reaches the matching models. Ask what the matching model optimises for, for a sub processor list covering both the client and worker sides, and for retention on location and call recordings.

Operational scale is well established (Home Instead's 1,000+ locations, 100,000 caregivers). AI specific outcome claims (retention, shift fill, cost reduction) are described by leadership in interviews but not published with methodology or independent validation.

The client side of this axis is addressed. The worker side, which is where the artificial intelligence actually operates, is not.

On clients, the company publishes a notice of privacy practices naming access controls limited to authorised employees, caregivers and business associates, confidentiality training for staff and partners, physical safeguards including secure offices and locked storage, contractual obligations on business associates, and breach notification. That is a real stewardship statement and it is graded on the health privacy axis.

On workers, three things are visible and none is explained. The care professional application uses a third party mapping interface for geolocation, disclosed in the terms of service by reference to that provider's own terms, with nothing on precision, retention, whether collection is bounded to a shift or continuous, or who inside the organisation can view a worker's location history. The privacy policy states that where data is collected in connection with employment it is also subject to employment policies, handbooks and agreements, which control in the event of any conflict, so the published document is not the operative instrument for the workforce and the operative one is not public. And the company states that telephone conversations may be recorded and monitored for quality assurance and training.

None of that is unusual for a workforce platform. What is missing is any statement of limits: what the location data is used for beyond dispatch, whether it feeds the matching or performance systems, and how long it is kept.

Considerably stronger than expected for a non medical home care business, and with one structural feature a licensing agency must understand before signing.

The company publishes a client notice of privacy practices in its own name, with a named privacy officer, a postal address and a dedicated contact address. It commits expressly to the health privacy rule, states the uses and disclosures it makes, names access controls, confidentiality training and physical safeguards, states that business associates are bound by contract, and commits to breach notification. A notice of privacy practices is the instrument of a covered entity rather than of a vendor, and publishing one at all puts this record ahead of most of the category.

The structural feature is this. The notice states that it applies to the franchise business and this company jointly, as the client's care provider. The company is therefore not positioning itself as a business associate supplying software to an agency. It is positioning as a joint provider of the care. An agency expecting an ordinary vendor relationship governed by a business associate agreement is entering something different, in which responsibility for the notice, for individual rights requests and for breach obligations is shared rather than delegated.

The role is also not the same everywhere. The company's own privacy policy states that for Ontario residents it acts solely as a service provider and directs privacy enquiries to the relevant franchise. Joint care provider in one country, service provider in another. Establish in writing which role applies to your arrangement, because the obligations that follow are different.

Held at B rather than A because no business associate agreement availability statement was located for agencies licensing the platform outside the franchise network, and because the notice does not identify which legal entity within the group signs what.

No SOC 2, ISO 27001, HITRUST or equivalent attestation was located and there is no trust centre.

The published notice of privacy practices does name administrative and physical safeguards, including access controls, confidentiality training, secure offices and locked storage, and it references delivering care through digital platforms supporting secure messaging. That is more than many comparable companies state, and it is self asserted rather than examined by anyone external.

The reason the absence matters more here than the grade alone conveys is what this platform holds. A clinical vendor holds information about a person's health. This one holds, for a large population of predominantly elderly people who often live alone, where they live, which hours of which days they are alone, what their physical and cognitive limitations are, who is scheduled to attend them, and in many operational models how a caregiver gains entry to the home. A compromise of that combination is a physical safety exposure and not only a privacy one, and it concerns a population selected for vulnerability. The scale is not small: the group reports a network in the order of a thousand locations and roughly a hundred thousand care professionals.

Ask for an attestation and its scope. Ask specifically how entry and access information is stored, who can retrieve it, and whether it is available to a caregiver outside the window of a scheduled visit.

A scoping determination and it closes cleanly. This is non medical home care: personal care, companionship and assistance with daily living rather than skilled nursing or medical treatment. The software matches caregivers to clients and schedules shifts. Neither the service nor the software makes a diagnostic or treatment claim about an identified person, so no device pathway attaches and none is claimed.

What governs this business sits elsewhere and a buyer should assess it there. State home care agency licensing determines who may operate and under what supervision, and varies substantially between states, which the company itself identifies as a principal difficulty for the agencies it serves. Labour and employment law is directly engaged, because the company states publicly that it is the employer of the care professionals working in the network. Franchise regulation applies to the franchise side of the group. And where care is funded by a public payer, electronic visit verification requirements attach to personal care services.

None of those regimes is the device regulator, and reading this axis as reassurance about the others would be a mistake.

The bias question here is not clinical, and it is more direct than in almost any other record in this index.

The platform's artificial intelligence matches care professionals to clients and fills shifts, and company leadership describes modelling caregiver motivations across many dimensions. The company also states on its own site that it is the employer of the local care professionals. Put those together and the system is an employer's automated tool allocating work to its own employees. Shift allocation determines who works and therefore who earns. That is an automated employment decision, and several jurisdictions now regulate such tools specifically, with requirements that can include independent bias auditing and notice to affected workers.

Nothing responsive was located. There is no description of what the matching model takes into account, no disparate impact testing, no bias audit, no notice that an automated system contributes to allocation, and no appeal route. The composition of this workforce is what makes the question pressing rather than theoretical: home care in the United States is staffed disproportionately by women, and by immigrant and minority workers, which is precisely the population an unaudited allocation system would affect first.

A second question sits on the client side and is specific to this industry. Home care staffing has a long documented problem of clients expressing preferences about a caregiver's characteristics. A matching system has to do something with such a preference: honour it, ignore it, or filter it out. Whichever it does is a governance decision with legal consequences for the employer, and nothing states which it is.

What would move this row is narrow: publish what the matching system considers, whether allocation outcomes have been tested across worker groups, and how client preferences that would be unlawful as employment criteria are handled.

Two passes located no accuracy figure, no evaluation methodology, no published limitations and no warranty, indemnity or remediation commitment, and the affected party here is not a patient, which is why this record reads differently from the rest of the lane. The models perform caregiver to client matching and scheduling, so the person the system decides about is a worker, and the terms governing that person are explicitly not the published ones.

The privacy policy states that where data is collected in connection with employment it is also subject to employment policies, handbooks and agreements, which control in the event of any conflict. So the document a reader can obtain is subordinated by its own text to instruments that are not public, and the operative rules for the workforce cannot be read by anyone outside. Three practices are visible and none is bounded.

The care professional application uses a third party mapping interface for geolocation, with nothing on precision, retention, whether collection is limited to a shift or continuous, or who inside the organisation can view a worker's location history. Telephone conversations may be recorded and monitored. And nothing states whether location or call data feeds the matching or performance systems that decide who gets offered work. None of this is unusual for a workforce platform.

What is missing is any statement of limits, and a caregiver has no route to see what the system holds about them or to contest a matching decision. Ask what the matching model optimises for, and what the employment instruments actually say.

A scoping determination on the clinical question, and an unanswered domain equivalent, which is why this does not sit higher.

Non medical home care operates largely outside the clinical record. No electronic health record integration is claimed and grading the product against clinical interoperability depth would misdescribe it.

The equivalent question is whether the platform connects to the systems a home care agency actually runs, and it matters here more than usual because the company describes taking over billing, scheduling, staffing and back office functions for its partner agencies. That is either deep integration with an agency's existing systems or wholesale replacement of them, and nothing published says which. An agency evaluating the partnership needs to know whether its current scheduling, payroll and billing systems continue to operate, are fed by the platform, or are retired.

One integration surface is specific and checkable and nothing addresses it. Electronic visit verification is a mandated requirement for personal care services funded by public payers in the United States, with state by state implementation and designated aggregators. Any agency serving publicly funded clients needs to know how the platform satisfies it and whether it submits to the state aggregator directly. Ask for the list of named integrations, the export formats available, and the electronic visit verification position state by state.

Cloud delivered, with a mobile application used by care professionals in the field. Beyond that the record is thin.

The privacy policy does make a transfer statement, which is more than nothing: personal information may be stored or processed in the United States or in other jurisdictions where data privacy laws differ, and Canadian users are asked to consent to transfer outside Canada with contractual safeguards said to apply. That is a transfer disclosure rather than a residency commitment, and it names no region, no data centre location and no option for a customer to constrain either.

Nothing was located on tenancy, on whether partner agencies are logically separated from one another, on retention periods, or on subprocessors. The one subprocessor visible anywhere is a third party mapping provider used for geolocation in the care professional application, disclosed in the terms of service by reference to that provider's own terms rather than in a subprocessor list.

For an agency partner the question that follows from the commercial model is worth asking directly. Where the platform takes over billing, scheduling and back office operations, the agency's own client and workforce records are being processed in the company's environment rather than the agency's. Establish where that environment is, who else is in it, and what the agency receives back if the partnership ends.

Converted from Not Rated, resolving one of the index's deliberately held rows. The specific question held open was whether the commercial structure described in this record is vendor published. It is not, and that is what settles the grade.

The structure itself is well established and consistently described: partner agencies pay a negotiated share of their revenue in exchange for the company taking over caregiver recruiting, onboarding and training, staffing, scheduling, billing and other back office functions, with the agency retaining client acquisition, assessment, contracting and licensure. That account appears repeatedly in trade press over several years and is attributed to named company executives in interviews and written statements.

What does not exist is any published commercial statement from the company itself. Its own network pages describe benefits to clients and to care professionals rather than terms to agencies. No percentage, no range, no floor or minimum, no contract term, no notice period and no rate card was located anywhere the company publishes. A prospective agency partner cannot form any estimate of what the arrangement would cost without entering a negotiation.

A structure known through journalism is not the same as terms published by the vendor, and this index grades the second. That distinction is the whole of the difference here, so if the company published the range and term this row would move immediately.

One related point worth raising in that negotiation, because it is a cost the headline share does not capture. Under this model the partner agency's operations run on the company's systems, so the switching cost at the end of the arrangement is a function of what the agency can take back with it. Ask what client records, caregiver records and scheduling history are returned, in what format, and on what timescale.

Scope is unambiguous: in home care for aging adults, delivered through licensed agencies and the Home Instead franchise network. Narrow, clearly stated coverage, which this axis rewards.