Upside

The AI is genuine and specifically described: acuity stratification scoring member risk, case summarization, and predictive matching against a curated housing inventory, with the company explicitly stating it uses a purpose built matching framework rather than general purpose language models. Held back because the delivered service is human: licensed social workers and housing specialists place and retain members, and the company describes its own model as human led and AI accelerated. Remove the AI and a slower version of the service still functions.

Oversight is structural by design, since a licensed social worker or housing specialist executes every placement and the AI prioritizes and summarizes rather than acts. The vendor states the intent is to remove repetitive tasks rather than replace the human side.

Held back from A because how acuity scores influence which members receive scarce housing resources is not documented, and a prioritization model allocating limited housing is a consequential decision that warrants published governance.

More technically specific than most records in this index, and unusually it includes a statement about what the company has chosen not to use.

What is public: three named artificial intelligence functions, acuity stratification, case summarisation, and predictive housing matching; a proprietary curated affordable housing inventory database spanning public and non public sources with real time vacancy tracking; and an explicit statement that the company deliberately avoids general purpose language models in favour of a purpose built framework for social determinants matching.

That last claim is what lifts this above the category norm. Almost every vendor in this index describes the capability it has added; very few describe an architectural choice they have declined to make and why. In this domain the reasoning is defensible on its face, since matching a person to a housing unit is a constrained optimisation over structured inventory and eligibility rules rather than a language task, and a generative system would introduce fabrication risk into a process where an invented vacancy or eligibility criterion has immediate consequences for someone in crisis.

Held at B rather than A because nothing beneath the claim is documented: no method for the matching, no accuracy or coverage figures for the inventory database, no description of how acuity is stratified or on what inputs, no versioning, and no evaluation. The inventory database in particular is the asset the whole model rests on, and its provenance, refresh cadence and completeness are undescribed.

Ask how matching is performed and evaluated, and how the inventory is sourced and kept current.

No stewardship framework, retention position or training use statement was located, and no party in the chain is named, while the data here is sensitive in ways health data usually is not. Housing instability status is not merely private: knowing that a person is facing eviction, is homeless, is couch surfing, or has been assessed as unable to live independently is information that can affect their tenancy applications, their credit, their employment, their custody arrangements and their standing with family.

It travels badly and it is difficult to withdraw once known. The employer channel is where that matters most, and the company's own evidence makes the point sharper than any outside observation could. It cites research that workers who lose their homes are substantially more likely to lose their jobs, and it sells a housing benefit to self insured employers.

So the company knows that housing loss predicts job loss, and it is offering employers a service that surfaces which of their employees are in housing crisis. What the employer learns, at what granularity, and with what separation between the benefit administrator and the employer's own management is the single most important question on this record, and nothing addresses it.

The housing network adds a third dimension, since placing a member requires disclosing information to landlords and housing providers who are not covered by health privacy rules at all. Ask what the employer sees, what reaches housing providers, retention, and whether member data trains the matching models.

Outcome claims are specific and tied to a defined window rather than asserted generally: more than half of enrolled members stabilized within 90 days, high enrollment velocity, and a reported return on investment within 12 months for risk bearing organizations. Adoption is independently corroborated at more than 17 health plans across 10 states including four of the largest US payers, with a named Medicaid contract from UnitedHealthcare of New Jersey. Held back from A because the figures are vendor reported without published methodology or a control comparison, and housing interventions are difficult to attribute cleanly against secular trends.

No stewardship framework, retention position or training use statement was located, and the data here is sensitive in ways health data usually is not.

Housing instability status is not merely private. Knowing that a person is facing eviction, is homeless, is couch surfing, or has been assessed as unable to live independently is information that can affect their tenancy applications, their credit, their employment, their custody arrangements and their standing with family. It travels badly and it is difficult to withdraw once known.

The employer channel is where that matters most, and the company's own evidence makes the point sharper than any observation from outside could. It cites research that workers who lose their homes are eleven to twenty two percentage points more likely to lose their jobs, and it sells a housing benefit to self insured employers covering rental and mortgage support, deposit assistance and elder transition planning. So the company knows that housing loss predicts job loss, and it is offering employers a service that surfaces which of their employees are in housing crisis. What the employer learns, at what granularity, and with what separation between the benefit administrator and the employer's management is the single most important question on this record, and nothing addresses it.

The housing network adds a third party dimension: placing a member requires disclosing information to landlords and housing providers who are not covered by health privacy rules at all.

Ask what the employer sees, what is disclosed to housing providers, retention, and whether member data trains the matching models.

No business associate statement, availability, scope or contracting entity was located, and this company sits across at least three different relationships without publishing a position on any of them.

With health plans it receives member information to identify housing instability and coordinate placement, which is business associate territory in the ordinary way. With state Medicaid housing supports contracts, the arrangement carries programme specific privacy and reporting obligations layered on top. And with self insured employers the analysis is different again: an employer sponsored benefit may sit inside the group health plan, which keeps it within health privacy rules and imposes strict limits on what reaches the plan sponsor, or it may be structured as an employer provided service outside the plan, which does not. Which structure applies determines whether an employee's housing crisis is protected health information or simply information their employer's vendor holds.

A fourth party is unavoidable and unregulated by these rules. Securing a placement means disclosing information about a person to landlords, housing providers and community organisations, none of whom is a covered entity or business associate. Whatever governs those disclosures, it is not the health privacy framework, and the member's consent is doing the work.

Ask which entity contracts for each line of business, whether the employer benefit sits inside or outside the group health plan, what authorisation supports disclosure to housing providers, and what the plan sponsor receives.

No SOC 2, HITRUST, ISO 27001 or equivalent attestation was located, and no trust centre or security page was found.

Context is worth stating. This is a company founded in 2020 that raised a twenty million dollar Series A in mid 2026, so a full attestation programme is a real cost at its stage, and the grade records what a counterparty can verify rather than asserting that controls are absent.

What raises the stakes above the company's size is who it contracts with and what it holds. It reports partnerships with more than seventeen health plans including four of the largest national payers, plus a named state Medicaid housing supports contract, which means large regulated counterparties have already conducted their own diligence. It also holds a combination of information that is unusual: health plan member data, housing instability status, assessments of whether a person can live independently, and placement records tying named individuals to physical addresses. A breach here discloses where vulnerable people live, which is a physical safety exposure rather than only a privacy one, and this index has recorded the same concern for home care and monitoring vendors.

The workforce dimension matters too, since care guides work in the field with member information on devices outside any office.

Ask what independent examination exists or is planned, how field staff access is controlled, how placement addresses are protected, and what security terms bind housing partners.

A scoping determination that closes cleanly, and the regimes that do apply are unusual enough to be worth naming precisely, because none of them is device regulation.

Nothing here is a medical device. The company delivers housing navigation, placement and tenancy support through licensed social workers and housing specialists, with software supporting triage, summarisation and matching. No diagnostic or therapeutic claim is made and none would be appropriate.

What governs instead is a stack this index has not previously had cause to describe. Social work licensure applies to the care guides and varies by state, as does supervision. Where the company delivers Medicaid housing supports under a named state contract, the terms of that programme and its state authority govern eligibility, service definitions and documentation. Landlord and tenant law governs the placements themselves and differs by jurisdiction and by unit type. And fair housing law is the one most directly engaged by the technology: it prohibits discrimination in housing on protected characteristics, it recognises disparate impact, and a system that ranks people against scarce inventory is squarely the kind of tool that has attracted scrutiny in tenant screening and advertising.

That last point deserves emphasis because it is the regulatory exposure a health care buyer is least likely to check for. A model allocating housing sits under housing law, not health law, and the two have different enforcers.

Ask what fair housing review the matching has had, and how state Medicaid contract obligations are evidenced.

No governance framework, evaluation methodology or bias analysis was located, and this vendor's models make decisions of a kind almost nothing else in this index makes.

The first is allocation of a rationed good. The company describes predictive housing matching against a curated database of public and non public affordable housing inventory, with real time vacancy tracking and placements secured within days. Affordable housing is scarce in every market it operates in, so a model that ranks members against available units is not surfacing an option, it is determining who gets housed and who waits. Whoever the model ranks first receives a unit that then does not exist for anyone else. Nothing published states what the matching optimises for, whether speed of placement, likelihood of tenancy success, cost, or member need, and those produce different queues.

The second is more consequential still and the company states it plainly. It describes an intervention triggered when a care guide, under clinical supervision, determines that independent living is no longer viable for a member's safety and longevity. That is a judgement about a person's capacity to live independently, reached by a contractor to their insurer, and it changes where that person lives. Nothing published describes what evidence supports it, what the member's voice is in it, whether they can contest it, or what role automated acuity stratification plays in reaching it.

Ask what the matching optimises, the demographic distribution of placements, and the process and appeal route for a viability determination.

This record does something almost nothing else in the index does: it states an architectural choice the company declined to make, and why. Three artificial intelligence functions are named covering acuity stratification, case summarisation and predictive housing matching, alongside a curated affordable housing inventory with vacancy tracking, and the company states explicitly that it deliberately avoids general purpose language models in favour of a purpose built framework for social determinants matching.

Almost every vendor here describes the capability it has added; describing a capability declined tells a reader far more, because it reveals how the company reasons about where the technology fits. The reasoning is defensible on its face: matching a person to a housing unit is a constrained optimisation over structured inventory and eligibility rules rather than a language task, and a generative system would introduce fabrication risk into a process where an invented vacancy or eligibility criterion has immediate consequences for someone already in crisis.

Held below the top grade because nothing beneath the claim is documented: no method for the matching, no accuracy or coverage figures for the inventory, no description of how acuity is stratified or on what inputs, no versioning, no evaluation, and no warranty, indemnity or remediation commitment. The inventory is the asset the whole model rests on and its provenance, refresh cadence and completeness are undescribed, and a stale vacancy is a wasted day for someone with nowhere to sleep. Ask how matching is evaluated and how the inventory is kept current.

A partial scoping determination, and the domain equivalent is asserted rather than described.

This is not a clinical product and no electronic health record integration would be expected. The buyer is a health plan or an employer, so the systems that matter are care management platforms, member eligibility files and claims systems rather than charts.

On that equivalent the company makes a general claim: care guides work closely with a plan's existing teams, housing plans are aligned with other social drivers and health strategies, and the service is described as easy to plug into current workflows with minimal lift. It also offers real time reporting, performance dashboards, member engagement data and return on investment metrics. Those describe outputs and working relationships rather than integration.

No named care management platform, interoperability standard, data exchange specification or eligibility file format was located, and nothing states how members are identified for outreach in the first place, which is the inbound question that matters most. Whether the plan pushes a cohort, whether the company screens against claims or eligibility data it receives, or whether risk stratification runs on plan supplied data determines both the integration and the privacy analysis.

One interoperability question is specific to this vendor and has no clinical analogue: the housing inventory database aggregates public and non public sources, and how that data is obtained and refreshed is an integration problem of its own.

Ask how members are identified and what data the plan supplies, what systems are integrated in production, and how housing inventory is sourced.

No hosting location, region, tenancy model, retention schedule or subprocessor list was located.

The operating shape is nonetheless partly visible and it is a services business with a software layer rather than a platform. Care guides work in the field across ten states. A technology layer performs stratification, summarisation and matching. A proprietary inventory database aggregates housing availability from public and non public sources. Health plan partners supply member data. Housing providers and community organisations receive placement information. Employers form a separate customer channel with its own data flows.

Two questions follow that are specific rather than generic. Tenancy separation matters because the company works with more than seventeen health plans including four of the largest national payers, which compete directly, and member level housing and utilisation information is commercially sensitive between them. And retention matters unusually here because a housing record is durable in a way a clinical episode is not: a placement history ties a named person to a sequence of addresses, and that record retains its sensitivity for as long as the person is alive.

A third concerns the field workforce, since staff working outside an office carry member information on devices in circumstances the company controls less tightly.

Ask where the platform and member data are held, how plan partners are separated, the retention schedule for placement and assessment records, and what applies to data held on field devices.

The commercial model is partially disclosed: contracts with health plans and employers, with the company describing outcome accountability and return on investment framing rather than a fee schedule. Buyers should establish whether pricing is per engaged member, per placement, or at risk against outcomes, since those structures allocate risk very differently for a service with variable housing costs. No rate card published.

Narrow and precisely stated: housing instability as a social determinant, across Medicaid, Medicare Advantage, D-SNP, commercial, and employer sponsored populations, spanning the continuum from crisis intervention to long term tenancy support. The company makes no clinical claims, which is the correct scope for what it does.