pareIT

The automated analysis is the entire product and there is no services business or workflow suite underneath it. Held at B rather than A because part of the extraction stack is licensed rather than owned: the platform runs on Amazon Web Services and uses Amazon Comprehend Medical, with the company's own patent pending models, trained on more than a million medicolegal data points, layered on top. That is materially more than reselling a third party service, and materially less than owning the medical language models outright, which is what an A describes on this axis. The grade describes the mechanism, not the quality.

No oversight mechanism is described. There is no confidence signal, no threshold, no abstention behaviour, no sampling or human verification step, and no statement of what a user must check before relying on an output that the company markets as litigation ready. The only thing performing that function is a liability disclaimer stating the service offers speed and accuracy to the best of our abilities subject to the terms of service. A disclaimer transfers risk to the user; it does not constitute oversight, and this index has consistently declined to grade a design philosophy or a caveat as an oversight mechanism. The assistant answering free text questions against the file widens the surface further, since the user composes the query and nothing the vendor validated covers it.

Two disclosures earn this grade and both are rare. First, pareIT answers the whose model is it question that this index puts to every vendor and that almost nobody answers: it names Amazon Web Services as its infrastructure and Amazon Comprehend Medical as part of its medical extraction stack, and states its own models are patent pending and trained on more than a million medicolegal data points. Naming the third parties in the stack is the same candour this index credited in OrbDoc, and it lets a buyer reason about where records are processed. Second, its stated accuracy position at launch was PARITY with manual medical record summaries rather than superiority to them, which is a restrained and more credible claim in a market whose competitors routinely claim to beat human reviewers. Held at B because no accuracy figure, methodology or evaluation supporting the parity claim was located, and because the site's own language hedges to speed and accuracy to the best of our abilities.

Nothing published establishes benefit or accuracy. No customer is named anywhere, no case study carries a quantified result, no independent evaluation exists, and a company data provider records no institutional or angel investors, so the company appears to be self funded and has no funding round to serve as external validation. The founder's domain credibility as a retained biomechanics expert is real but is not evidence about the software. The throughput claims also do not reconcile across sources: a thousand pages in as little as two hours at launch in 2022 against up to four thousand pages per hour in later material, roughly an eightfold difference with no explanation of what changed or how either was measured.

HIPAA compliant data handling, encryption, access controls, multi factor authentication and regular penetration testing are attributed to the platform, though by a third party software directory rather than confirmed on the vendor's own current pages, which could not be retrieved. No retention period, no statement on whether customer files are used to train or improve models, and no de identification posture was located, and marketing language about continuous learning and adaptation points toward model improvement without bounding it. One consequence of the stack disclosure belongs here rather than only as a transparency credit: because extraction runs partly through a third party cloud medical language service, records pass through a processor other than the vendor, and the resulting agreement chain should be established explicitly.

HIPAA compliance is claimed, with no business associate agreement terms published, which is the standard middle rung. Two qualifications. The claim was located through a third party directory and the vendor's own current pages could not be retrieved to confirm it. And the market nuance recorded across this group applies with particular force here, because the case types span personal injury, medical malpractice, criminal matters and workers compensation, where records are typically obtained under authorisation, subpoena or discovery rather than under HIPAA treatment provisions. Establish which instrument governs the records in your own matters before treating a HIPAA claim as the relevant assurance.

SOC 2 is listed for the company alongside encryption, access controls, multi factor authentication and regular penetration testing, but the source is a third party software directory rather than the vendor, the report type is not specified, and no trust centre, security page or scope statement was located. Type I versus Type II is the whole assurance question. Graded C for that combination of unspecified type and unverifiable source. Ask which report exists, over what period, who issued it, and request it directly.

No FDA clearance or device authorisation was located and none is expected, since the output informs a legal or claims decision rather than a diagnosis or treatment decision. The regulatory and professional exposure sits elsewhere: rules of professional conduct governing an attorney's supervision of non lawyer and automated work product, evidentiary and disclosure obligations where a generated chronology informs expert testimony or a demand package, and discovery rules on how the output itself may be treated.

The grade describes disclosure, and there are real credits alongside the gaps. pareIT names its model stack, states limitations openly in its terms, and made a parity rather than superiority accuracy claim, all of which are more restrained than the market norm. Against that, no fairness, subgroup or demographic performance disclosure of any kind was located. And the consequence surface is the widest in this group, because the case types include CRIMINAL MATTERS alongside personal injury and medical malpractice. A missed or mischaracterised entry in a medical chronology carries different weight in a criminal proceeding than in a claims file, and nothing published addresses what verification the product expects before its output reaches that setting. The combination of a free or near free offering, litigation ready framing and a best of our abilities disclaimer is worth weighing together rather than separately.

Graded against what this product is rather than penalised for a mismatch. There is no EHR integration and none should be expected for a medicolegal buyer. What exists is one meaningful named integration, Clio, the dominant legal practice management platform, which puts the output where the case file already lives, plus document ingestion covering scanned and handwritten records and a beta capability extracting ICD and CPT codes. Graded C because Clio is the only named integration, no API availability was confirmed, and no standards based exchange is described.

Web based hosted software with document upload, running on Amazon Web Services, which is named explicitly and is more than most vendors in this category disclose. No region, residency commitment, customer hosted option or data location choice was located, and because part of the processing runs through a third party cloud medical language service, the path a record takes is longer than a single vendor boundary. Establish where files are processed and stored, and for how long.

More commercial openness than anything else in this group, and a trajectory that needs explaining. At launch the company published a relative position, roughly a third of what such services typically cost. Later material describes transparent pricing per page with no monthly fees or subscriptions, which is a genuinely clear model and unusual in a market that hides behind demo requests. The current site markets a free offering. Held at B rather than A because no actual figures could be retrieved, the vendor's own pages disallow automated access, and the movement from a third of market price, to per page, to free is unexplained. A buyer should establish what the free tier covers, what triggers a charge, and whether output rights or data use differ between free and paid usage.

The broadest case type coverage in this group. Personal injury and workers compensation are shared with the other insurance and legal products here, but medical malpractice and criminal matters are unique to this record, and the company also addresses insurers and health technology buyers alongside law firms. Those are genuinely different workflows: a malpractice chronology turns on standard of care and a criminal one on entirely different questions from a damages timeline. Graded B rather than A because the breadth is asserted through market positioning rather than demonstrated through case type specific behaviour, and no instrument level depth was located in any of them.