Superinsight

AI Centrality

How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.

The models are the product and the billing unit. Every output is model-generated, from chronologies and narrative summaries through residual functional capacity assessments to research queries, and charging is by credit consumed per report, stated as roughly two credits a report, so a customer pays for inference and nothing else. The company goes further than most in describing the pipeline as fully automated, stating that no human reviewers see the records and that only its AI processes them. Remove the models and there is an upload form. Checked 4 September 2026.

Citation Accuracy and Hallucination Disclosure

Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.

Accuracy is asserted through outcome claims and no grounding mechanism is described on any surface read. The claims are specific in form but unsupported in substance: 3x more evidence found than manual review, 70 per cent time savings on the home page against up to 90 per cent in the FAQ on the same page, and 80 per cent cost reduction, none with a test set, a method, a sample or a date. The two time-saving figures contradict each other and are named here because a reader who finds that unaided trusts the rest less. Nothing describes how an output traces back to the page it came from: no citation, no source link and no verification view appears on the home, pricing or legal surfaces. The agreement points both ways, with clause 2.4 committing to promptly correct material defects that materially impact the accuracy of reports while clause 6 disclaims any warranty of accuracy at all. Documentation pages on building and managing reports were not opened in this pass.

Autonomy and Oversight Model

What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.

Full automation is published as a selling point and no oversight mechanism is described alongside it. The process is stated in three steps, upload, AI analyses, receive a report, and the FAQ says expressly that it requires no human review of sensitive medical data. That is framed as a privacy advantage rather than an autonomy claim, but it is the same fact: nothing between the record set and the report but the model. Against that, the terms place the whole burden on the user, with clause 3.1 stating the user assumes all risks associated with use of content including any reliance on its accuracy, and clause 6 disclaiming accuracy outright. What is absent is any described review surface: no verification view, no confidence signal, no threshold at which the system defers, and no statement that a lawyer must check the output before use. Documentation pages covering report building, report management and research insight were not opened and are named so the limit is visible.

Operational and Outcome Evidence

Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.

Real deployment evidence with unusually complete attribution, short of dated measurement. Five testimonials each carry a photograph, a full name, a role, an organisation and a practice area: Katie Reed of McMahan Law Firm on personal injury, Scott Haider of Schneider Law Firm on disability and personal injury, Diane C. Haar of Hawaii Disability Legal Services on disability, and Paul Bunn of Veterans Outreach Ministries and Dawud D.K. Shillingford of Shillingford Consulting Group on VA disability. One carries a hard figure tied to the named organisation: Bunn states the product saved about 300,000 dollars a year and allowed him to replace three registered nurses and two nurse practitioners while maintaining review quality, which is a specific and checkable operational claim rather than a satisfaction quote. The company separately claims more than 500 attorneys and hundreds of firms. What holds this at B is that nothing is dated, no case study is published, and the corporate percentages are not attached to any named customer.

Privilege and Confidentiality Posture

How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.

Substantive published commitments, with the strongest element being one most vendors leave silent. The HIPAA Notice of Privacy Practices, effective 12 January 2026 and published rather than held behind a contract, states that Superinsight does not share personal information including protected health information with vendors performing services for it, naming artificial intelligence and large language model companies and cloud and database integration services, except as required by law or with explicit authorisation. That is a direct answer to what a third-party model provider may see, and it is a rarer disclosure than a subprocessor list. It sits alongside a marketing commitment that no human reviewers access the records. The content licence is narrow: clause 3.2 authorises use of content solely for the purpose of providing information back to the user, with all other rights retained. What holds this at B is the rest of the picture. Nothing addresses privilege or work product, nothing describes segregation between customers or matters, and no retention period exists anywhere, with clause 3.1 stating instead that user content may be deleted at any time without prior notice.

UPL and Professional Responsibility Posture

Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point.

No position on the advice line was located anywhere, and the audience makes that a live gap rather than a formal one. The terms of use, the privacy policy and the HIPAA notice were all read in full on 4 September 2026 and none states that output is not legal advice, not medical advice or not a substitute for professional judgement, and no rule of professional conduct, bar authority or accreditation regime is named. The product is sold to lawyers, to independent and qualified medical examiners producing expert evaluations, and to non-lawyer representatives: two of the five published testimonials come from organisations that are consulting and outreach practices rather than law firms, working veterans disability claims. Representation in VA and Social Security matters is governed by its own accreditation and fee rules, and a product generating residual functional capacity assessments and evidence arguments for those claims engages them directly. Nothing published addresses any of this.

AI Governance and Bias Disclosure

Published governance over model behaviour: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.

A governance standard is invoked without a framework behind it. The home page states the platform is ISO 42001-aligned, alongside HIPAA compliance, and links a Trust Cloud portal. Alignment is the operative word and it is doing a lot of work: ISO 42001 is an AI management system standard that can be certified against by an accredited body, and a claim of alignment is a self-assessment rather than an audited certification, which is why this does not reach the band that a genuine certification would. Nothing behind the claim is published: no management system scope, no named owner accountable for model behaviour, no evaluation regime, no testing results, and nothing at all on bias, which matters on a product that decides which clinical findings support a disability or causation argument. The Trust Cloud portal at superinsight.trustshare.com was not opened in this pass and is named rather than credited.

AI Safety and Data Stewardship

Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.

Parts of the set are published and the parts that govern the records after processing are not. What exists is real: the HIPAA notice commits to breach notification in accordance with HIPAA requirements, gives rights to inspect, amend, restrict and obtain an accounting of disclosures, and states that no AI or LLM vendor receives protected health information. Access control is described as a one-time passcode sent to the user's phone plus encryption and firewalls. Three things are missing or dated. No retention period is published anywhere, and the terms instead reserve the right to delete user content at any time without prior notice while disclaiming any backup obligation. No deletion route or certification is offered. And the security language is stale, with the privacy policy describing secure socket layer technology and carrying a last-updated date of 21 August 2023, three years behind the HIPAA notice it sits alongside.

AI Liability and Recourse

What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.

Liability is addressed only through a standard limitation clause that disclaims the exposure the product creates, and the cap is the striking number. Clause 7 limits liability for any damages arising from or related to the terms to a maximum of two hundred and fifty US dollars, stating expressly that more than one claim does not enlarge the limit. Clause 6 provides the app as-is with all warranties disclaimed including accuracy. Clause 4 runs the indemnity in one direction only, from user to company. Clause 5.3 goes further than most, with the user releasing and forever discharging the company from every claim relating to the app including personal injuries and death, and California residents waiving Civil Code section 1542. Two published items point the other way and are recorded: clause 2.4 commits the company to maintain the app in good working order and promptly correct material defects that materially impact the delivery or accuracy of reports, and the home page offers a no-risk money-back guarantee. Both are service commitments rather than allocations of loss. No insurance position was located.

Practice Systems Integration Depth

How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.

An integration capability is published in the documentation index without any practice system named on a Superinsight surface. The documentation site carries an Integration page under Manage Case, which establishes that the product connects to something, and the app supports case sharing, groups and organisation management. What is absent is the detail an implementer needs: no case management system, document management system, intake platform, medical record retrieval service or e-signature product is named anywhere on the home page, the pricing page or the legal documents, and there is no integrations page or API reference on the marketing site. Records arrive by upload in the published three-step workflow. The Integration documentation page was not opened in this pass and is named here so the limit is visible rather than presented as a settled absence; it is the cheapest available upgrade on this record. Checked 4 September 2026.

Deployment Model and Data Residency

Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.

Neither the tenancy model nor the region is stated. No hosting provider, cloud platform, country, region or data centre is named on any surface read, which is unusual: most records at least name an infrastructure supplier. Nothing describes whether the platform is single or multi-tenant, and no dedicated, private or isolated option appears at any of the six published pricing tiers. The HIPAA notice and the privacy policy are both silent on where protected health information is stored or processed, and the privacy policy contains no international transfer section at all. The one adjacent statement is the terms' export control clause referring to US export laws. Searched the home page, the pricing page, the terms of use, the privacy policy, the HIPAA notice and the site navigation on 4 September 2026.

Security Certifications and Trust Center

Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.

Compliance is claimed on two fronts and neither is an independent attestation, though a trust portal exists. HIPAA compliance is claimed repeatedly, but HIPAA has no certification scheme, so the claim is self-attested by construction. ISO 42001 is claimed as alignment rather than certification, which is the vendor's own careful wording and should be read as a self-assessment. No SOC 2 of any type is claimed, no auditor or certifying body is named anywhere, and no report date, scope statement or penetration testing appears. What keeps this off the floor is a real access route: a Trust Cloud portal is publicly linked from both the home page and the FAQ at superinsight.trustshare.com. It was not opened in this pass, so its contents and whether it is self-serve or gated could not be established, and the lower tier is assumed for that reason and stated here. Two further badges are recorded as what they are rather than as security credentials: NVIDIA Inception membership is a startup programme and the CIO Applications listing is a magazine award.

Model Supply Chain Disclosure

Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.

The vendor states who does not see the data and never states what does. The HIPAA notice is unusually direct on the negative half, committing that protected health information is not shared with vendors performing services for Superinsight, naming artificial intelligence and large language model companies expressly, and the privacy policy repeats the same commitment for personal information. Read together with the claim that no human reviewers access records, the implication is that processing runs on models the company controls, but that is inference rather than disclosure: no model is named, no version is given, no provider is identified, no hosting location is stated, and no commitment to notify customers if any of it changes was located. A buyer can therefore establish that no named third party receives their records while being unable to establish what actually reads them.

Commercial Transparency

Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.

A buyer can price this completely, compare tiers and start without speaking to anyone. Six options are published with figures: one-time credits at 80 dollars with a one-credit minimum and no commitment, then Starter at 250 dollars a month for 4 credits, Pro at 500 for 12, Grow at 1,250 for 36, Scale at 2,500 for 80 and Enterprise at 5,000 for 200, with yearly billing stated to save 17 per cent and to deliver thirteen months of credits upfront. The unit is defined rather than assumed, with a report stated to average about two credits, and a comparison table gives the effective cost per credit at each tier from 80 dollars down to 23. What changes between tiers is itemised across six rows covering templates, priority support, a dedicated account executive and unlimited research queries. Expiry rules are published and specific: monthly credits expire at the end of the cycle, yearly credits roll over, one-time credits expire 30 days after purchase. The first case is free, no credit card is required, and a three-question tool recommends a tier.

Firm and Practice Coverage

Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.

Practice coverage is the most granular in this lane and the boundary is left open. Eight practice areas each carry a dedicated page with a described analytical approach rather than a label: Social Security Disability with residual functional capacity assessments, personal injury plaintiff with causation and damages, personal injury defence with pre-existing conditions and treatment inconsistencies, veteran disability with service connection, workers' compensation with causation and impairment, mass tort with qualifying injuries and exposure documentation and Camp Lejeune named specifically, medical malpractice with standard of care deviations, and medical examiner services for physicians and IMEs. Firm size is addressed indirectly but usefully through the pricing tiers, which are sized by monthly case volume from one or two up to fifty to a hundred or more. Fifty state-specific pages are published. What is missing is the limit: in-house and government use are not addressed, and rather than stating where the product stops, a Custom Solutions page offers to build for any practice area not listed.