Quench

Extraction, summarisation, chronology construction and natural language interrogation of the record are the product, with no services organisation or prior software underneath. The founder states the company built its own engineering rather than wrapping a general purpose model, framing the work as proprietary and grounded in the team's clinical and data science background. That assertion could not be independently verified and is recorded rather than credited, but the product itself is unambiguously model driven.

Oversight here is achieved by scope limitation rather than by adding a check, which for this use case is the stronger approach because it removes the inference surface instead of monitoring it. Quench states publicly that its AI does not perform legal tasks such as citing cases or drawing conclusions, and that the system exists to organise documents, provide a chronology and help a reviewer find relevant information. Evaluation, findings and conclusions remain with the professional.

Supporting that, the interface returns documented answers, findings are saved as the reviewer works rather than generated wholesale at the end, and the chronologies are framed as defensible.

Held at B because nothing is quantified: no accuracy rate for the chronology or the missing record detection, no confidence signal, no abstention behaviour, and no specification of what documented means at the level of an individual answer. Ask what a documented answer cites, and at what granularity.

An unusual inversion of the normal gap. Most vendors in this index say nothing about what they run on. This one says what it is NOT, with the founder stating it is not ChatGPT and not just a wrapper and describing sophisticated proprietary engineering, without saying what the stack actually is. A negative claim is not a disclosure.

No model or model family is named, no accuracy figure exists for extraction, chronology construction or the question answering interface, and no evaluation methodology was located. AskQuench is described as returning documented answers, which suggests citation to source, but the granularity is unspecified.

More operational detail than most of this group, and the same lifecycle gaps. The company describes encrypted transmission and storage, access controls restricting patient information to authorised personnel, background checks on all employees before employment, security training at onboarding and annually, a named privacy officer directing security, and a named cloud host.

Several of those are rarely stated at all, and the personnel controls are the ones worth crediting specifically: pre employment screening and recurring training address the insider path, which is where a document review business is most exposed and which almost every vendor in this index leaves unmentioned while describing encryption at length. Naming the privacy officer gives an outside party someone to write to. What is absent is the lifecycle.

No retention period, no statement on whether uploaded records are used to train or improve models, no de identification posture, and no sub processor list. The data subject position matches the rest of this group and bears on how much the vendor should publish: the person whose records are reviewed is an examinee, a claimant or an opposing party who has no relationship with the vendor, did not choose it, and has no route to ask what was held or for how long. Ask for retention once a matter closes, the training position in contract language, and what a data subject can obtain about a review of their own file.

Credentialed voices, no independent evidence. Two named clinical advisers hold verifiable senior positions in occupational health at the Cleveland Clinic and at a Department of Veterans Affairs healthcare system, which is meaningful domain credibility for a medicolegal product.

One caution that applies generally and applies here: the testimonials carried on the site are attributed to those same clinical advisers, so they are endorsements from people affiliated with the company rather than independent customer references, and should not be read as the latter. No customer organisation is named, no funding is disclosed, no accuracy or time saving figure is published, and no independent evaluation was located.

More operational detail than most of this group and the same gaps. The company describes encrypted transmission and storage, access controls restricting PHI to authorised personnel, background checks on all employees before employment, security training at onboarding and annually, a named Privacy Officer directing security, and Amazon Web Services hosting. Those are real controls and several of them, particularly the background checks and the annual training, are rarely stated at all.

What is absent is the lifecycle: no retention period, no statement on whether uploaded records are used to train or improve models, and no de identification posture. The data subject position also matches the rest of this group, since the person whose records are reviewed is an examinee, a claimant or an opposing party who has no relationship with the vendor and did not choose it.

HIPAA compliance is claimed with more supporting specificity than the norm, covering encryption, access control and a named Privacy Officer, though no business associate agreement terms are published. The market nuance recorded across this group applies with force: independent medical examination, qualified medical evaluation, workers compensation and litigation work frequently sit outside the HIPAA covered entity framework, and records typically arrive under authorisation, subpoena or discovery rather than under treatment provisions. Claiming HIPAA compliance in that context is a credit rather than a requirement. Establish which regime governs the records in your own matters.

No third party attestation of any kind was located: no SOC 2, no ISO 27001, no HITRUST, and no trust centre. That is what holds this at C.

Against it sits something this index rarely finds and which deserves crediting: a published vulnerability disclosure process, with a named contact address, stated expectations for what a report should contain, a commitment to verify rapidly and to send periodic status updates until a problem is fixed. Very few vendors anywhere in this index publish a route for a security researcher to reach them at all, and doing so is a genuine maturity signal even without an attestation behind it. Named infrastructure hosting on Amazon Web Services is also more than most disclose.

Ask whether a SOC 2 is planned or underway.

No FDA clearance or device authorisation was located and none is expected, since the output informs a legal, claims or evaluation process rather than a diagnosis or treatment decision. The relevant exposure sits elsewhere: evidentiary and discovery rules where a chronology supports expert testimony, professional obligations governing an expert witness's reliance on assistive tools, and state workers compensation and independent evaluation regulation.

The grade describes disclosure, and one item on this record is the strongest governance behaviour found anywhere in the medicolegal group. Quench publicly declines a capability that would sell: it states its AI does not cite cases or draw conclusions, and confines itself to organising, chronologising and locating information.

In a market where a direct competitor in this index markets inference as its central differentiator, describing its product as drawing conclusions about whether a condition progressed, that is a deliberate limitation on a revenue generating surface and it is the correct one for a use case where an inference error cannot be caught by checking a source. It also publishes a vulnerability disclosure process and names its clinical advisers rather than hiding behind anonymous endorsement.

Held at C because nothing is measured: no accuracy figure, no fairness, subgroup or demographic performance disclosure of any kind, and no published evaluation. Publishing any accuracy or subgroup data would move this to B immediately, and the company is closer to deserving it than most of this group.

An unusual inversion of the normal gap sits on this record and it is worth naming as a category. Most vendors in this index say nothing about what they run on. This one says what it is not, with the founder stating that the product is not a consumer chat product and not just a wrapper, and describing sophisticated proprietary engineering, without saying what the stack actually is.

A negative claim is not a disclosure: it forecloses one reading while establishing nothing, it cannot be checked, and it invites a buyer to supply a flattering inference about what sits behind it. It also tends to appear precisely where the positive statement would be easy to make, which is what makes it worth flagging rather than ignoring. Beyond it there is nothing.

No model or model family is named, no accuracy figure exists for extraction, chronology construction or the question answering interface, no evaluation methodology was located, and no warranty, indemnity or remediation commitment attaches. The chronology is the component to press, because building a timeline from a disordered file is where a document review product creates its own version of events, and an omitted or misdated entry changes the narrative a decision rests on.

The interface is described as returning documented answers, which suggests citation to source, and the granularity is unspecified. Ask what the stack is, for accuracy on chronology construction, and the citation granularity.

Graded against what this product is rather than penalised for a mismatch. There is no EHR integration and none should be expected, since the users are expert witnesses, evaluators, law firms and insurers who do not operate one. Ingestion is document upload, handling large disorganised PDF sets, and review happens inside the platform with what the company describes as Adobe like precision, which keeps the reviewer in one tool rather than switching between a viewer and a summary. No API, case management integration or standards based exchange was located, so the product sits as an island in the reviewer's workflow.

Hosted software running on Amazon Web Services, which is named explicitly and is more than most vendors in this category disclose, with the company citing AWS data centre physical access controls as part of its security posture. No region, residency commitment, customer hosted option or data location choice was located. Graded C because the hosting arrangement is clear and the residency picture is absent.

No price, tier or pricing mechanism was located.

The company does offer a self service free trial with an open sign up, which is a lower friction commercial route than the demo request that gates most of this category and lets a reviewer test the product on their own records before any conversation. That is credited, but it is access rather than price. Publishing the tiers behind that trial would be straightforward for a product sold to individual practitioners and small firms, and would move this record up.

Broad across matter types and reviewer roles rather than clinical specialties, which is the right measure for a medicolegal product. Named coverage spans workers compensation claims, standard of care evaluation, injury from defective products, litigation, claims disputes and underwriting, and the user set includes physicians and nurses acting as expert witnesses, independent medical examiners, qualified medical evaluators, law firms and insurance companies.

The occupational health emphasis is genuine rather than incidental, reflected in both named clinical advisers and in the workers compensation and independent evaluation focus. Graded B rather than A because depth within any single matter type is asserted rather than demonstrated.