Claim Health

The model does the work end to end: extracting referrals from fax, email, and portal, identifying coverage and documentation risk inside the EMR before submission, and automating prior authorization through renewal. The stated long term goal is a self driving revenue cycle.

The stated division is sensible: routine work is automated while exceptions surface to human billers who make the important decisions, which puts the human where judgment is actually required. Held back from A because the threshold separating routine from exception is not published, and that threshold is the entire control in a system designed to reduce human touches.

No model architecture, evaluation methodology or accuracy figure was located.

The published evidence is operational rather than technical: a billing process reported to move from three days a week to roughly two hours at an early customer, and company revenue growth of thirty times in under a year. Those describe adoption and time saved, not how often the system is right.

The number that matters most follows from the product's own design. Routine work is automated and exceptions surface to human billers, so the threshold separating routine from exception is the entire control. Nothing published states where it sits, who sets it, how often a claim classified as routine is later denied, or what feedback loop adjusts the boundary. For a product whose value proposition is reducing human touches, the rate at which an untouched claim goes wrong is the honest measure of whether that is safe.

No retention period, statement on whether customer data trains or improves models, or de identification posture was located, and no model or hosting arrangement was named. The ingest surface is wide for an early company: referrals arrive by fax, email and portal and are automatically extracted, and the platform reads documentation and coverage detail inside the medical record, so faxed and emailed referrals are unstructured clinical documents and the material held includes narrative about a patient's condition and care needs rather than claim fields alone.

Two questions are worth asking now rather than later, and the timing argument is the point. They are cheap to answer while a company is small and expensive to unwind once the corpus exists, because a retention rule adopted at the start costs nothing and a purge designed three years in has to reckon with everything already accumulated and everything already learned from it.

What is retained from an extracted referral once the case is routed, given that the routing decision is what the customer needed rather than the document. And does material from one agency inform models serving another, which is commercially live as well as legally relevant here, because post acute agencies compete directly for referrals from the same hospitals and referral patterns are exactly what a competitor would want. Ask both, in writing, before the corpus is large.

Very early with no external validation retrieved. Founded 2025, seed funded January 2026. Notably, C suite executives at large post acute providers participated in the round, which is a modest signal of operator conviction, but no customer counts, named references, or measured outcomes were available.

No retention period, no statement on whether customer data is used to train or improve models, and no de identification posture was located.

The ingest surface is wide for an early company. Referrals arrive by fax, email and portal and are automatically extracted, and the platform reads documentation and coverage detail inside the electronic medical record. Faxed and emailed referrals are unstructured clinical documents, so the material held includes narrative about a patient's condition and care needs rather than claim fields alone.

Two questions are worth asking now rather than later, because they are cheap to answer while a company is small and expensive to unwind once the corpus exists. What is retained from an extracted referral after the case is routed, and does material from one agency inform models serving another. Post acute agencies compete directly for referrals from the same hospitals, which makes the second question commercially live as well as legally relevant.

No HIPAA compliance statement and no business associate agreement terms were located. Business associate status is structurally certain, since the platform reads the clinical record and processes referrals and claims on behalf of provider agencies.

The buyer profile is what makes the absence consequential rather than routine. Home health, hospice and home care agencies are typically small organisations without a privacy office or a mature vendor assessment function. They are the buyers least able to extract compliance detail through negotiation and the most helped by finding it published. This index has recorded the same pattern for community oncology practices, and the reasoning carries.

One flow deserves specific attention. Referrals arrive from hospitals and physician practices that are separate covered entities and are not the vendor's customers, so patient information reaches the platform from organisations with no agreement with it. Establish the authority under which that intake happens.

No SOC 2, HITRUST or ISO 27001 attestation was located across two differently phrased searches, and no trust centre or security page was found.

One correction belongs in the record because it would be easy to get wrong. A third party article about the company's founder contains language about end to end encryption and compliant infrastructure, but that text is the publishing site's own boilerplate about its services rather than a statement about this vendor, and no company source repeats it. It is not credited.

Stage context, so the grade reads fairly: the company was founded in 2025, came through an accelerator in 2025 and raised seed funding in January 2026. A completed attestation would be unusual at that point and its absence is not a judgement on the team. It remains the first thing a post acute provider should ask for, because these are small operators with limited vendor review capacity handling patient data across fax, email and portal intake.

No FDA pathway applies and none is claimed. Revenue operations work on referral, eligibility, authorisation and billing data has no diagnostic or treatment decision surface.

Graded C because post acute care carries one of the heaviest audit regimes in Medicare and no position is published on any of it. Home health and hospice are subject to targeted probe and educate reviews, unified programme integrity contractor audits, and pre claim review requirements in several states. Hospice carries its own election statement, face to face encounter and aggregate cap requirements. Payment in both home health and skilled nursing is driven by clinical assessment instruments rather than procedure codes, so the assessment is the claim.

That last point is why this matters more than for an acute revenue cycle vendor. A platform validating assessment data before submission is operating directly on the instrument that determines payment, and the provider carries False Claims Act exposure for what is submitted. Establish what the system may change without a human confirming it, and whether any assessment field can be altered by automation.

No AI governance framework, model monitoring disclosure or bias evaluation was located.

The mechanism here is unusually direct and deserves stating carefully, because the same capability runs in both directions. The platform identifies coverage and documentation risk before a claim is submitted. The company's stated purpose for that is to help agencies accept more patients, and it describes providers currently having to turn away patients they cannot afford to serve. That is a real good and it should be credited as the intent.

The same model, pointed the other way, predicts which referrals are likely to be reimbursed. An agency under margin pressure that receives a risk score at the point of referral has a tool for declining patients more precisely. The patients flagged as risky will tend to be those with complicated coverage, incomplete records, or histories that make documentation harder, which is not the same as those who need care least.

Nothing published addresses whether risk scores are reported by patient characteristics, whether acceptance rates are monitored after deployment, or whether a score can be surfaced at intake at all. Those are the questions that decide which direction it runs.

No model architecture, evaluation methodology, accuracy figure or warranty, indemnity or remediation commitment was located. The published evidence is operational rather than technical, covering a billing process reported to move from three days a week to roughly two hours at an early customer and rapid company revenue growth, and those describe adoption and time saved rather than how often the system is right. The number that matters most follows from the product's own design.

Routine work is automated and exceptions surface to human billers, so the threshold separating routine from exception is the entire control: everything classified as routine passes without a person looking, and everything the threshold misclassifies is by definition unreviewed. Nothing published states where that line sits, who sets it, how often a claim classified as routine is later denied, or what feedback loop adjusts the boundary.

For a product whose value proposition is reducing human touches, the rate at which an untouched claim goes wrong is the honest measure of whether that reduction is safe, and it is also the figure a buyer needs in order to size the staffing they can actually release.

The company is early and the absence is unremarkable at this stage, which is precisely why the question is worth asking now: the threshold is being tuned against live claims either way, and someone should be recording the result. Ask where the threshold sits, who owns it, and the denial rate on untouched claims.

Integration depth is real and specific to a segment most vendors skip. The platform works inside the post acute electronic medical record to identify documentation and coverage risk, centralises referrals arriving by fax, email and portal into one workflow with automatic extraction, and connects to clearinghouses and payer systems for submission and reconciliation.

The detail that evidences genuine depth is field level assessment validation. Published material shows the platform enforcing conditional logic on specific assessment data elements, which means it operates on the structured instrument that drives post acute payment rather than on claim headers. That is a harder integration than reading a claim file and it is the right place to catch an error.

Held at B rather than A because no post acute electronic medical record platforms are named and no integration method is described. The post acute vendor landscape is concentrated among a handful of systems, so naming them would be straightforward and would tell a buyer immediately whether their stack is supported.

No hosting provider, region, tenancy model or data residency commitment was located, and no subprocessor list is published.

Two subprocessor questions are specific to this product. Referrals arrive by fax and email, and fax carriage is almost always operated by a third party, so patient documents pass through infrastructure the agency has not assessed before reaching the platform. And voice agents for insurer information requests are described as in development, which will add a telephony and speech processing layer when it ships.

Ask for the subprocessor list now and ask to be notified when it changes, since the stack is still being assembled and an early customer will inherit whatever is added later.

No public rate card. Contact the vendor. The commercial unit is undisclosed and materially important here: subscription versus percent of collections allocates risk very differently for a revenue recovery product, and buyers should establish which applies.

Narrow, deliberate, and underserved: home health, hospice, home care, and nursing homes. Post acute revenue cycle differs materially from acute and ambulatory in documentation requirements and authorization rules, and the large RCM vendors largely do not build for it. Confining scope to that segment is the right call and is stated plainly.