pareIT

The model output is the product. pareIT sells medical chronologies, full medical analyses, executive summaries, red flag analysis, answers from its Paige assistant and document OCR, all generated from the uploaded file, priced by page processed. There is no services business or workflow suite underneath: remove the models and nothing is left to sell. Whether part of the extraction is licensed does not change that.

The home page sells completeness and correctness: "PareIT extracts every relevant fact and cites its source - so nothing gets left behind", "Expert-level output every time", and a featured customer quote says the tool "never misses a thing". Those are completeness and correctness universals, and they grade D whatever the review mechanism, because an omitted fact cannot be seen in a chronology and a buyer told nothing is left behind will not reread the file.

What the product does offer is a citation to the source page for every finding, which lets a reviewer check what was extracted but not what was missed. No confidence signal, threshold or required verification step is described for chronologies, the Red Flag Analysis or the Paige assistant, even though the output is sold as ready for depositions, hearings and patient care decisions. The terms say pareIT "does not guarantee that the services will be performed error-free".

The approach and a supplier have been named, without version or update practice. The 2022 launch announcement described patent pending machine learning models trained on more than a million medicolegal data points, running with Amazon Comprehend Medical on AWS; the current site describes the engine as patented and proprietary and not a repurposed off-the-shelf tool, and describes its steps (sorting and deduplicating documents, building a timeline, extracting diagnoses, treatments and handwritten entries, citing each finding). The terms describe the service as AI powered medical record review. No model version, update practice or evaluation is published, and whether the 2022 stack still applies is not stated.

The model layer was named, in 2022, and the current site says something different. The company's June 2022 launch announcement, as carried in its developer's case study, said the platform was powered by AWS and Amazon Comprehend Medical, with the company's own patent pending models on top.

The current home page instead describes a patented, proprietary engine that "isn't a repurposed off-the-shelf tool", and the terms bar disclosure of medical records to any third party except as required by law, which sits oddly beside a cloud medical language service processing them. No subprocessor list, retention terms or training terms for any party are published. A provider was named, but nothing current says whether it is still in use.

Figures with no stated provenance and no named customer. The home page claims 75 percent less review time, 75 percent cost savings, a 98 percent accuracy rate against manual review and more than 4,000 pages processed an hour, without saying where any was measured, on what files or against which reviewers.

The only customer voice is a quote from an initialled personal injury attorney in Texas, with no firm named, and a line that the product is trusted by legal and medical professionals nationwide. The 2022 launch announcement claimed a thousand pages in about two hours and a third of the usual cost; the current figures are several times faster with no explanation of what changed or how either was measured.

Categorical commitments are published, without a retention schedule or an explicit training statement. The privacy policy says medical records are used solely to provide the services and will not be shared with any third party without the user's consent except as required by law; the terms repeat that pareIT will not disclose medical records to any third party other than as required by law, and say the company has no obligation to keep data and it may be irretrievably deleted 60 days after the service is delivered.

That is a permission to delete, not a commitment to, so how long records and outputs actually persist is unstated, and nothing says in terms whether uploaded records are used to train or improve the models, which the company's launch material described as trained on a large medicolegal dataset. De-identification is not addressed.

HIPAA status is stated and supported by documents that reach the service, with no agreement published. The home page calls the platform HIPAA-compliant and HIPAA and SOC 2 compliant. The privacy policy has a medical records section (records accessed with the user's authorization, used solely to provide the services, not shared without consent), and the software as a service agreement treats uploaded medical records as confidential and bars disclosure to third parties except as required by law.

No business associate agreement is referenced. The role question depends on the buyer. Law firms and insurers usually obtain records under authorization, subpoena or discovery, where HIPAA may not govern pareIT's handling at all; the site now also sells chart review, prior authorization and clinical summaries to physicians and nurse reviewers, and for a covered entity customer pareIT would be a business associate.

The company presents a SOC 2 as its own: the home page carries a SOC 2 Certified badge and says the platform is HIPAA and SOC 2 compliant. The report type, auditor, period and scope are not stated, and no trust center or security page is linked; the privacy policy says only that information is stored on secure servers behind firewalls. Type 1 against Type 2 is the first question.

No device claim is made, and the product now reaches closer to care than it did. For legal and insurance buyers the output informs cases and claims, not treatment. But the site also sells to physicians and nurse reviewers, offering clinical summaries, problem extraction, medication reconciliation and lab analysis, and describes the output as ready for patient care decisions. The company states no regulatory position.

Each finding cites its source in records the customer uploaded, so the basis is reviewable, and nothing published suggests the product diagnoses or recommends treatment; the healthcare offering is where the exposure sits.

Nothing is published on how model behavior is governed or tested. The home page says every output is fully cited and auditable and that the technology is patented, proprietary and built by practitioners, but there is no evaluation method, error taxonomy, subgroup result or governance process behind the 98 percent accuracy claim, in the privacy policy, the terms or the pricing page.

The exposure is wide because the buyers are: chronologies and red flag analyses feed depositions, settlements and malpractice cases, and the site now also offers chart review, medication reconciliation and lab analysis to clinicians for patient care decisions. Extraction quality varies with the records themselves, handwriting, scan quality and the systems they came from.

Mechanisms let a user check an output, and nothing stands behind it. Every finding cites its source page. The software as a service agreement warrants only that the system performs in all material respects as described, requires written notice within five business days of delivery, limits the remedy to a fee credit for that service, states that the services are not guaranteed error-free, caps liability at the amount paid for the service in question, provides trial use as is, and has the customer indemnify pareIT for claims arising from the services. The 98 percent accuracy figure has no method or denominator, so it is not a published error rate. A warranty that the product works as described is not attached to the chronology itself.

Named systems, one directional. The home page describes native integrations with Clio, the legal practice management platform, and with ChartSquad, a medical record retrieval service, and the enterprise tier adds JSON and API output. The site also says the product is EHR-compatible and EHR-integrated for healthcare customers and accepts EHR exports as input, but names no record system and shows no production connection.

A single hosted web application with location implied rather than committed. Users upload files to the web application; the privacy policy says information is stored on secure servers behind firewalls, and the company is a California LLC contracting under California law. The only named host is AWS, in the 2022 launch announcement. No region, residency commitment or customer hosted option is stated, and whether record text passes through an outside medical language service today is unresolved.

Published tiers with figures, a stated unit of charge and a route to start without a sales conversation. The pricing page offers a first case free with no card, then pay as you go at $0.10 a page; Boutique at $100 a month (or $1,000 a year) with 2,000 pages and $0.07 overage; Power Player at $199 a month (or $1,990 a year) with 5,000 pages and $0.05 overage; and a custom enterprise tier for 25,000 to 100,000 or more pages a month.

Add-ons are priced too: full medical analysis with red flags at $50 or $25 a case, and assistant question packs at $5 for ten. One inconsistency to settle: the home page puts the cost at about $300 per 1,000 pages, three times the pay as you go rate. Also confirm whether pages are counted before or after duplicates are removed.

pareIT serves two named buyer groups and offers a general document product. For legal professionals: medical chronologies for personal injury and medical malpractice cases, executive summaries for depositions, and the Clio integration. For healthcare professionals: chart review, prior authorization, clinical summaries, problem extraction, medication reconciliation and lab analysis. Document OCR is offered across legal, medical, insurance and real estate work.

The pricing tiers are written for law firms, and the healthcare offering is described only at headline level, so depth there is asserted rather than shown. Earlier material also named criminal and workers compensation cases, which the current home and pricing pages do not mention.