Reveleer

A record retrieval and abstraction business rebuilt around AI rather than an AI native company, and the history matters for sizing it. Founded in 2009 as a medical record retrieval operation, the company now applies natural language processing and machine learning through what it calls an Evidence Validation Engine that parses retrieved charts and populates abstraction fields for coder review.

The retrieval infrastructure, meaning the relationships and workflows that get charts out of provider offices at volume, remains a substantial non AI asset and is arguably what customers cannot easily replace.

The AI populates and the human confirms. The engine automates chart retrieval, parses records, and pre fills abstraction fields, with certified coders reviewing before submission, and output is described as maintaining document linked evidence so a reviewer can trace each abstracted element back to source. That structure is appropriate given the end product is a regulatory submission to CMS where the plan bears audit liability for every code.

The pipeline is described functionally, covering automated retrieval, record parsing, abstraction field population, and document linked evidence for audit defense, and third party analysis characterizes the underlying technology as natural language processing and AI. No model architecture, accuracy figures, or validation methodology were located, which is a meaningful gap for a vendor whose output feeds RADV submissions.

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 position on retention, de identification or training use was found. What makes this record consequential is an architectural change that enlarges what the unnamed chain holds.

The company states it manages, reads and processes more than 1.2 billion pages of clinical data annually on behalf of more than fifty five health plans, which is among the largest protected health information surfaces of any vendor in this index and is the company's own figure rather than an inference.

A clinical data repository introduced in 2026 then shifts the model from episodic retrieval, where records are pulled for a purpose and used for it, to a persistent enriched member level record held once and reused across risk adjustment, quality and audit. The company frames that as reducing duplicate retrieval and provider friction, which is a genuine benefit to everyone including the provider organisations repeatedly asked for the same chart.

It also converts a transient holding into a durable one. A vendor assembling longitudinal member records across dozens of plans holds something categorically different from a retrieval queue, and the retention, access, secondary use and deletion questions all change with it while none is addressed. Ask what the repository retains, for how long, who can query it across plans, and whether it trains models.

Operational rather than clinical, appropriately, and the strongest evidence is longevity and scale rather than published metrics. The company has operated in medical record retrieval since 2009 and serves health plans across Medicare Advantage, ACA Marketplace, and Medicaid, with third party analysis noting suitability for high volume chart review projects and backlog clearing. It has raised substantial capital including a 65 million dollar round in 2024. No published accuracy figures, customer outcome studies, or independent evaluation were located.

Converted from Not Rated. The surface is now quantified by the company itself, and the architecture has just changed in a way that enlarges it further.

The scale is stated plainly: managing, reading and processing over 1.2 billion pages of clinical data annually on behalf of more than 55 health plans. That is among the largest protected health information surfaces of any vendor in this index, and it is the company's own figure rather than an inference.

The architectural change is the more consequential development and it postdates the earlier review. A clinical data repository launched in 2026 shifts the model from episodic retrieval, where records are pulled for a specific purpose and used for that purpose, to a persistent, enriched, member level clinical record held once and reused across risk adjustment, quality and audit. The company frames this correctly as reducing duplicate retrieval and provider friction, which is a real benefit to everyone including the provider organisations repeatedly asked for the same chart.

It also converts a transient holding into a durable one. A vendor that assembles longitudinal member records across dozens of plans is holding something categorically different from a retrieval queue, and the retention, access, secondary use and deletion questions all change with it. None of them is addressed publicly.

Nothing was retrieved on retention periods, de identification, whether member records enrich model development, or what is returned at contract end.

Ask what the repository retains, for how long, and whether it trains models.

Converted from Not Rated. No published position was located, and the company's operating model makes it a three party question rather than a two party one.

No business associate agreement, addendum, role statement, subcontractor flow down, breach notification timetable or review cadence was retrieved. What exists is a positioning phrase, that the platform is designed to support customers' compliance needs, which describes an intention rather than an obligation.

The structure is what deserves attention. The core operation is retrieving complete medical records from provider organisations on behalf of health plans. The health plan is the covered entity that contracts, and this vendor is its business associate. But the records are physically obtained from provider organisations that are separately covered entities and are not the customer. Those providers hand over charts to a company they did not select, under a permitted disclosure they must satisfy themselves about, and with no agreement of their own governing what happens next.

That asymmetry is inherent to retrospective risk adjustment rather than a failing of this vendor, and it is worth naming because the provider bears effort and exposure while the plan holds the contract. The 2026 repository sharpens it, since a chart handed over for one purpose now persists for reuse across others.

Business associate status here is long established, the company having operated as a retrieval business since 2009, so agreements plainly exist. None is public.

Ask for the agreement, the subcontractor terms, and what providers are told about retention and reuse.

Corrected from Not Rated. A HITRUST position is claimed in the company's own materials, which the earlier review did not locate.

The company describes its newest platform as built on a HITRUST certified foundation, alongside governance controls it names specifically: audit traceability and role based access. HITRUST is the healthcare specific framework and the one plan procurement teams most often require, so holding it is substantive rather than decorative.

Two things hold this at B rather than higher, and both are precision issues this index tracks consistently.

The level is not named. HITRUST i1 is a fixed control set certified for one year; r2 is the tailored risk based two year certification assessed across far more controls. A buyer told only that a vendor is HITRUST certified would reasonably assume the higher tier. Peers elsewhere in this index name the level and are credited for it.

The phrasing is also ambiguous in a way that matters. Built on a certified foundation can mean the company holds the certification, or that the platform sits on infrastructure someone else certified. Those are materially different, and this index does not credit an infrastructure provider's certifications to the vendor running on top of them. Company material elsewhere suggests a direct certification, and a buyer should confirm which is meant.

No SOC 2 report of either type and no trust centre were located.

Ask for the certificate, the level, the scope of systems covered and the current expiry date.

No FDA pathway applies. The operative regime is CMS and NCQA, and the company's product scope maps directly onto it: HEDIS abstraction for quality reporting, risk adjustment coding, and RADV Independent Validation Audit submissions.

Supporting the IVA workflow specifically is notable, since that is the contractor led validation stage where a plan's documentation is tested against submitted codes, and building for it means the vendor is operating at the sharpest edge of the compliance surface.

Converted from Not Rated. The prior analysis identified the right technical risk, and there is a second, directional one that belongs alongside it.

No governance framework, bias evaluation or performance breakdown was located.

The technical risk is parsing variability. Retrieval spans thousands of provider organisations running wildly different record systems, and documentation quality varies enormously between a large integrated system and a small independent practice. A model that reads some formats and handwriting reliably and others poorly will find conditions unevenly, and the gap follows practice resourcing rather than patient health. That is the same mechanism this index has recorded elsewhere, where image or documentation quality acts as a proxy for how well funded the setting was.

The directional risk is the one specific to risk adjustment. Finding an additional hierarchical condition category raises the payment a plan receives, so the commercial gradient runs toward finding more conditions, and enforcement rather than the vendor is what constrains it. The company sells into both sides of that: coding to capture conditions, and audit tooling to defend the result when validation comes. Selling the capture and the defence is coherent and it is also worth a buyer noticing.

What would settle it is published evidence of calibration: the ratio of conditions added to conditions removed, and validation of suggested codes against an independent standard rather than against acceptance.

Ask for parsing accuracy by record source and format, and for the add to delete ratio on suggested codes.

Document linked evidence for audit defence is the mechanism and it is the reason this is not lower: an abstraction that points back to the page it came from can be checked by a reviewer and produced to an auditor, which is exactly what the output is for. Held at C because nothing measures the abstraction itself.

No accuracy figure, precision, recall or validation methodology was located, and no model architecture is described, so a buyer cannot establish how often a field is populated wrongly or left empty when the evidence was present. The consequence is more specific than for most products in this index.

This output feeds risk adjustment data validation submissions, which are federal audits where an unsupported code is recovered and a pattern of them is a compliance problem for the plan rather than for the vendor. Both error directions cost the plan: an abstraction that captures a condition the record does not support creates exposure, and one that misses a supported condition leaves revenue and, more importantly for the member, a condition uncaptured in the record used to plan their care.

Neither rate is published. No warranty, indemnity or remediation commitment was located. Ask for precision and recall on abstraction, the rate at which abstracted codes survive audit, and what the vendor commits to when an unsupported code is submitted.

The company's distinctive capability is breadth of retrieval across the provider landscape rather than depth of integration with any one system, which third party analysis describes as accelerated record collection at scale. That same analysis notes the integration lift is real, requiring coordination across data systems. No named EHR integrations or standards support were located.

Converted from Not Rated. No hosting provider, region, tenancy model, subprocessor list, backup posture or residency commitment was located.

The gap matters more here than for a typical software purchase because of what the platform now holds. Processing over 1.2 billion pages annually for more than 55 health plans, and since 2026 maintaining a persistent member level clinical repository rather than a transient retrieval queue, means the deployment question is where a very large longitudinal clinical corpus physically resides and how plans are separated within it.

Tenancy is the specific unknown to press. Competing health plans in the same markets are customers simultaneously, and their member populations overlap, since a person changes plans while their medical history does not. A repository built around a unified member level record across enterprise operations raises a reasonable question about whether member records are partitioned strictly by contracting plan, and how a record assembled under one plan's contract is treated if that member moves. Nothing published addresses it.

The retrieval side has its own footprint that is never described: obtaining charts from thousands of provider organisations implies connections, portals, credentials or on site collection, each of which is a distinct exposure.

Ask where the repository is hosted, how tenancy is separated between plans, what happens to a member record when the member changes plan, and what the return and deletion terms are at contract end.

Converted from Not Rated. No pricing, no basis and no structure is published, across a business that runs two economically different models side by side.

The company offers platform software and end to end managed services for record retrieval and review. Those are not variants of one price. Software typically prices per member, per plan or per module; a managed service prices per chart retrieved, per record reviewed or per full time equivalent replaced. A plan comparing this vendor against another may be comparing entirely different units without either party saying so.

A third possibility is specific to this category and worth raising directly. Risk adjustment vendors are sometimes compensated as a share of the incremental payment their work produces. If any part of the arrangement is contingent on conditions found, the commercial incentive points the same direction as the coding, and the party absorbing an over captured condition is the federal programme rather than the contracting plan. This index has made the same structural observation about contingency pricing on the payment integrity side, where the identical logic runs the other way.

Nothing published indicates whether contingency arrangements are offered, so the question is open rather than answered.

The stated value is quantified in operational terms, less retrieval cost, less provider friction, faster audit response, which is more concrete than most, and still not a price.

Ask for the pricing unit for each model, and whether any component is contingent on conditions captured.

Broad across the payer quality and risk surface, spanning quality improvement including HEDIS, risk adjustment, and member management, with program coverage across Medicare Advantage, ACA Marketplace, and Medicaid. The orientation is retrospective look back work rather than point of care support, which third party analysis identifies as the tradeoff against prospective competitors. Buyers are health plans and risk bearing provider organizations rather than clinicians.