ProVision

ProVision is a side letter and most favoured nation management platform built for the lawyers who run private investment fund formations. Its subject is the work that happens after a fund closes, when negotiated investor terms sit scattered across execution copies, spreadsheets and email threads. The platform takes uploaded side letters, limited partnership agreements and existing master side letters, extracts the provisions and investor-specific terms from them while preserving the original legal text, and assembles them into a single structured master side letter view that a team can search, compare and merge. From there it runs the MFN process end to end, generating investor-specific election forms in PDF and Word, distributing them with built-in anonymisation, and tracking responses back into one record of what each investor elected. It also holds LP comment memos and supports redlining provisions across investors to identify differences and apply consistent headings. The company is explicit that the platform does not require artificial intelligence to run, and that AI is applied where it adds value with guardrails intended to keep control of outputs with the lawyer. Deployment is single-tenant on Amazon Web Services with a dedicated instance per customer, role-based access control, AES-256 encryption at rest and TLS 1.2 or higher in transit, and configurable data residency; a trust centre describes this architecture and lists technical, compliance and legal documents available by request from the company's information security address. ProVision is made by Intelligent Legal Solutions (ILS) Limited, founded in 2024 and based on City Road in London, by co-founders Fergus Plant and Jack McCarthy, who worked in the investment funds teams at Goodwin and Proskauer, with Stefano Benigni. The company reports that in 2025 the platform processed more than 5,000 side letters across over 450 matters, supporting more than 150 billion dollars of capital raised across the United States, United Kingdom, European Union, Hong Kong, Singapore and the Middle East, and names Cleary Gottlieb, Fried Frank, Goodwin, Paul Weiss, Proskauer, Ropes & Gray and Willkie Farr & Gallagher among its customers.

Vendor siteLondon, United KingdomFounded 2024
Last verifiedSeptember 4, 2026

Capability grades

All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.

BB on AI CentralityThe models are the engine of a core capability, layered on a product that would still function without them as a document or workflow system.

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 drive a core capability and the vendor states plainly that the product does not depend on them. Extraction is the model's job: uploaded side letters, LPAs and master documents are read and their provisions and investor-specific terms structured automatically. But the support FAQ says in terms that the platform does not require AI to run and that the most advanced AI solutions are deployed only where they add value, with guardrails to retain control over outputs, and a customer's own published account describes AI-powered functionality as something added when appropriate according to each client's preferences and data security policies. So the AI is optional by design and per-client. What remains without it is real rather than residual: a structured repository of side letters and LP comment memos, MFN election form generation and response tracking, provision comparison and merging, versioning and an action record. That is a workflow and document system the models sit on top of, which is this band rather than the one above. Checked 4 September 2026.

Source: Vendor Published
CC on Citation Accuracy and Hallucination DisclosureAccuracy is asserted without measurement, or grounding is claimed while output cites sources the reader cannot open and verify.

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, real architectural controls exist, and nothing is measured. Two design commitments are published and both bear on grounding: extraction is stated to structure provisions while preserving the original legal text, and comparison is described as operating without breaking traceability, so an output can be traced to the source clause it came from. The support FAQ adds that fail-safes or guardrails are provided to prevent hallucinations. Against that, no accuracy figure, no test set, no evaluation, no error rate and no description of extraction method appears on any surface, and nothing states what proportion of provisions are extracted correctly or how a mis-extraction surfaces to a reviewer. The bottom band does not fire here, because this is not a bare hallucination claim standing alone: preservation of original text and traceability are genuine controls. A Data Security White Paper described as a technical deep dive into the architecture exists but is obtainable only by emailing the company. Searched the home page, the ProVision overview, the about page, the support FAQ and the trust centre on 4 September 2026.

Source: Vendor Published
BB on Autonomy and Oversight ModelA written commitment that the models work alongside a supervising lawyer, with real review surfaces, short of the full control structure: commonly the threshold at which the system stops or what happens after it is wrong.

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.

A clear written commitment that the lawyer stays in control, with real review surfaces and no described control structure. The about page states that the technology is there to assist rather than to second-guess professional judgement and that ProVision keeps the lawyer in control, and the overview repeats that structuring happens without compromising legal judgement. The review surface is concrete rather than asserted: extracted provisions and investor information are presented for review, MFN elections are defined by rules and parameters the team sets, and the home page describes clear visibility and control at every step with a record of every action, approval and document version. What is absent is the structure around it. No threshold is published at which the system acts without a reviewer, nothing describes what happens to extractions nobody approves, no confidence signal is surfaced against an individual provision, and the guardrails referred to in the support FAQ are named without being described.

Source: Vendor Published
BB on Operational and Outcome EvidenceReal deployment evidence with substance, short of full attribution or measurement: a named customer without figures, or figures without the named customer.

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.

Named customers at the top of the market and figures that do not attach to any of them. The customer list is specific and published by the company: Cleary Gottlieb, Fried Frank, Goodwin, Paul Weiss, Proskauer, Ropes & Gray and Willkie Farr & Gallagher, described as more than ten of the world's largest law firms. Two named individuals are quoted with their roles, Theresa Spartichino, Director of Practice Technology at Ropes & Gray, and Conan Hines, Director of Practice Innovation at Fried Frank, and Fried Frank has separately published its own announcement about deploying the platform, quoting Becky Zelenka, partner and co-head of its Private Equity Funds group. The volume figures are dated and substantial: in 2025 the platform is stated to have processed more than 5,000 side letters across over 450 matters, supporting more than 150 billion dollars of capital raised across six named markets. What holds this at B is that the two sets do not meet. No figure is attributed to any named firm, the quotes describe operational clarity and rapid adoption without measurement, and no case study is published.

Source: Vendor Published
BB on Privilege and Confidentiality PostureSubstantive published commitments on confidentiality and training use, short of the full picture: commonly silence on segregation between users or matters, or on what the underlying model provider may retain.

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.

The segregation limb is answered better than almost anything in this corpus and the rest of the picture is gated or unread. What is published is specific and checkable: ProVision is deployed in a single-tenant AWS environment logically segregated per customer, documents and extracted data sit in a dedicated instance and are stated never to be commingled with other clients' data, there is no shared document storage across customers, access runs on granular role-based control, and encryption is AES-256 at rest with TLS 1.2 or higher in transit. That is tenant-level separation documented at the level a law firm buyer requires. Three things hold it here. No privilege or work product treatment is named anywhere, which the top band requires as its own limb. The training question is addressed by a published FAQ entry, Do you use our data to train AI models, whose answer sits in a collapsed accordion this instrument could not expand, so the position could not be established. And a further claim needs recording rather than crediting: the trust centre describes a zero-knowledge framework under which ILS cannot access customer provisions or investor data, and end-to-end encryption at every stage, which sits in tension with a product that extracts and structures provisions server-side. Nothing published reconciles the two, and the white paper that would is available only by email.

Source: Vendor Published
CC on UPL and Professional Responsibility PostureA boilerplate disclaimer sits in the terms while the marketing describes the product in advice terms, or the intended audience is left ambiguous.

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.

Something real is published about professional judgement and nothing at all about the advice line. The about page states that the technology assists rather than second-guesses professional judgement and that ProVision keeps the lawyer in control, and the overview describes structuring side letters and LPAs without compromising legal judgement. The audience is unambiguous and narrow, being private investment funds lawyers at law firms and fund formation teams, so there is no consumer-facing surface and the disclosure limb of the top band does not bite. What is missing is the advice line itself: no statement that outputs are not legal advice, no requirement that a lawyer verify an extracted provision before relying on it, no rule of professional conduct, bar or regulator named, and no jurisdiction limit stated despite the company reporting work across six markets with different regimes. The document that would ordinarily carry a disclaimer is the Terms of Service, described on the trust centre as the governing agreement for using ILS platforms and obtainable only by emailing the company, so it could not be read. Neither this band nor the one below reads cleanly, since there is no boilerplate disclaimer to point at and the audience is not ambiguous; graded at the nearer and the reason stated.

Source: Vendor Published
DD on AI Governance and Bias DisclosureNo governance position published for a system whose output affects legal outcomes.

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.

No governance position was located. There is no responsible AI page, no principles statement, no named owner accountable for model behaviour, no evaluation or testing regime, and nothing at all on bias or on uneven extraction quality across document types, drafting styles or jurisdictions. The site inventory was taken from the navigation and footer across three pages on 4 September 2026 and runs to ProVision Overview, Trust Centre, Blog, Support, About ILS, Our Team, Careers, Press Announcements, Contact, Privacy Policy and Cookies Policy; none of them is a governance surface. The trust centre's Resource Vault lists eight items and every one is a security, privacy or contractual document rather than an AI governance artifact. The nearest published language is the support FAQ's reference to fail-safes and guardrails, which is a product assurance rather than a governance framework and is graded on the accuracy and autonomy rows instead of counted twice here. A description of the company as taking a lawyer-led and disciplined approach to AI comes from its lead investor rather than from the company.

Source: Operator Verified
CC on AI Safety and Data StewardshipA generic privacy policy covers the product without addressing what happens to documents and prompts after processing.

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.

Access control and encryption are published in detail and everything governing the data afterwards sits behind an email request. Published in the clear: AES-256 at rest, TLS 1.2 or higher in transit, single-tenant AWS deployment with logical segregation per customer, granular role-based access control, real-time monitoring, redundant backups and configurable data residency options. That is a real account of how the data is held. What is not published is the rest of the set. No retention period appears anywhere, no deletion route or certification is described, and no subprocessor is named on any readable surface. The documents that would answer each of those are listed on the trust centre and every one resolves to a mailto link at the company's information security address: a GDPR Overview stated to cover where data is hosted and which subprocessors are involved, a Data Processing Agreement, an Incident Response Overview and a Privacy Policy. Where a portal does not state whether its access flow fulfils without a sales conversation, the lower tier is assumed and that is the position here. A Privacy Policy also sits at a public footer URL and was not opened in this pass.

Source: Vendor Published
DD on AI Liability and RecourseNothing published on who bears the loss when the system is wrong.

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.

Nothing on the allocation of loss could be read, and the governing document is named rather than published. The trust centre lists a Terms of Service and describes it as the governing agreement for using ILS platforms, ensuring legal clarity and compliance, and its View link resolves to a mailto address at the company's information security team. The Data Processing Agreement sits behind the same gate. No indemnity, no cap, no carve-out, no warranty on output and no insurance position appears on any readable surface, and neither the home page, the overview, the about page, the support FAQ nor the trust centre body carries any liability language at all. This is a documented absence rather than an established silence: the vendor plainly has an agreement and states so, and it is not published in a form a buyer can read before entering a sales conversation. The grade records what a reader can establish from published material as located on 4 September 2026, and obtaining the Terms of Service would be applied as an amendment with its own date. It is the cheapest available upgrade on this record.

Source: Operator Verified
DD on Practice Systems Integration DepthNo integration into practice systems located, or the product stands alone and requires work to move to it.

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.

No integration into the systems legal work already lives in was located, and the product's own described workflow is upload and export. Documents enter by being uploaded, described as side letters, LPAs and related documents uploaded in minutes, and outputs leave as generated election forms in PDF and Word. Nothing connects to a document management system, and neither iManage nor NetDocuments nor SharePoint is named anywhere; no practice management, matter management, e-billing or fund administration system is named either; and there is no integrations page in the site navigation or footer, no API reference and no developer documentation. The one connection published is authentication rather than practice systems: the support FAQ confirms single sign-on with multi-factor authentication, or username and password with two-factor, or both. That governs how a user reaches the platform, not how a matter file reaches it. The absence carries weight on a product whose input is execution-version documents that in an AmLaw firm already sit in a document management system. Searched the home page, the ProVision overview, the about page, the support FAQ and the trust centre on 4 September 2026.

Source: Operator Verified
BB on Deployment Model and Data ResidencyDeployment model is stated clearly with partial residency detail, or residency is offered without the processing location being addressed.

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.

The tenancy model is stated clearly and unusually prominently, and the residency half is offered without detail. ProVision is published as deployed within a secure single-tenant AWS environment, logically segregated per customer, with each customer's documents and extracted data in a dedicated instance and no shared document storage across customers. The vendor makes the point comparatively, stating that most legal technology providers operate shared multi-tenant environments and that ProVision is different, so this is a positioning claim it has committed to rather than a passing mention. Amazon Web Services is named as the host. Residency is where it stops short: configurable data residency options are listed as a feature and no region is named anywhere, so a buyer knows a choice exists without knowing what the choices are. Nothing states where processing happens as distinct from where data is stored, which matters on a product whose extraction step may run elsewhere, and no model or inference location is published. A GDPR Overview stated to cover where data is hosted exists behind an email request.

Source: Vendor Published
BB on Security Certifications and Trust CenterCertification is real and stated, short of accessible evidence: a named standard without scope, date, or a way to obtain the report.

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.

A real trust centre and a scoped report, undercut by the vendor contradicting itself on whether it holds the certification. The trust centre states that a SOC 2 Type II report is independently issued and covers security, availability and processing integrity, which names the standard and its scope. But the same page's compliance section says the systems are aligned with SOC 2 Type II standards, and its summary strip reads SOC-aligned, while the support FAQ states the company is SOC 2 Type 2 certified. Alignment is self-assessment and certification is not, and the two cannot both be the position. The conflict is the finding and it is what prevents this reaching the top band, alongside three ordinary gaps: no auditor is named, no audit period or report date appears, and the report itself is reachable only by emailing the company's information security address, with the portal not stating whether that request fulfils without a sales conversation, so the lower tier is assumed. What keeps it above the band below is that these are not unsupported badges: a scoped, independently issued report is claimed to exist and a route to it is offered.

Source: Vendor Published
CC on Model Supply Chain DisclosureThe vendor refers to advanced or proprietary models without identifying what sits underneath.

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 refers to advanced AI without identifying anything underneath it. The support FAQ says the platform deploys the most advanced AI solutions only where they can add value, which is a characterisation of the models rather than a disclosure of them. No model is named, no provider is identified, no version is given, no processing location for inference is stated, and no commitment to notify customers when any of it changes was located. Amazon Web Services is named and is not counted here: naming a cloud host establishes where the vendor's platform sits, not whose model reads a side letter, and the same fact is graded on the deployment row. One route to the answer exists and is closed: the trust centre lists a GDPR Overview stated to cover which subprocessors are involved, and its link resolves to a mailto address. A buyer can therefore establish that AI is used on their investor documents and nothing whatever about whose it is, which is the substance of this band.

Source: Vendor Published
DD on Commercial TransparencyNo pricing information published at any level, including the unit of charge.

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.

No pricing information is published at any level, including the unit of charge. The site inventory taken from the navigation and footer on 4 September 2026 contains no pricing page, and the support page routes pricing explicitly to a sales conversation, listing a Sales contact for ProVision pricing, enterprise discussions and rollout planning behind a mailto link. Nothing states whether the platform is charged per matter, per fund, per side letter processed, per seat or as a firm-wide licence, and no tier structure, band, minimum, term or implementation cost appears. Two FAQ entries touch the commercial question, What is the ROI and What does a pilot look like, and both are collapsed accordions whose answers this instrument could not expand; neither is a pricing disclosure in any event. No pricing row is written, which is the correct outcome where the only thing published is an invitation to contact sales.

Source: Operator Verified
BB on Firm and Practice CoverageSegment and practice coverage is described with substance, short of the boundaries: what is supported is clear, what is not is left open.

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.

The practice is defined with unusual precision and the boundary is not stated. The product is addressed to private investment funds lawyers and fund formation teams, and the work is named at the level of the task rather than the practice group: side letters, limited partnership agreements, master side letter compilation, most favoured nation elections and LP comment memos, described as post-close fund workflows. Segment is evidenced by the named customer set, which is elite and international, and geographic reach is published as the United States, United Kingdom, European Union, Hong Kong, Singapore and the Middle East. The buyer is also identified on both sides of the relationship, with the platform described as serving funds and law firms alike. What is missing is where the product stops. No statement identifies fund types or structures it does not support, nothing addresses whether an in-house fund legal team can buy it directly as against through counsel, no firm size floor is given, and the roadmap language about expanding into a unified platform for fund formation teams leaves the current edge of the product undefined.

Source: Vendor Published

Legal Signals

What each signal means

A signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.

Confidentiality and Privilege

Client Data in Training

Can material a lawyer puts into this product be used to train a model?

Terms silent

No located term or policy addresses the question either way.

The vendor addresses this question directly and the answer could not be read. The trust centre and support pages both publish an FAQ entry headed Do you use our data to train AI models, so the position exists and is offered to buyers; the answers on both pages render as collapsed accordions that this instrument cannot expand. The retrieval ladder was run and partly succeeded on the same pages: search recovered the answers to other entries from the identical accordion set, including the statement that the platform does not require AI to run and deploys advanced AI only where it adds value with guardrails, and the confirmation of single sign-on with multi-factor authentication, but not this one. The two documents that would settle it are gated, with the Data Processing Agreement and the Terms of Service both reachable only by emailing the company's information security address. This row records what could be located as of 4 September 2026 and is expressly not a finding that ILS is silent on training or that it trains on customer content; the vendor publishes a heading this index could not open. Recovery of the answer or of either document would be applied as an amendment with its own date.

Source: Operator VerifiedAs of Sep 4, 2026

Prompt and Output Retention

How long does the product keep what a lawyer typed, and can that be set to zero?

Not addressed

No located public material states how long prompts and outputs are retained.

No retention period for uploaded documents, extracted provisions or generated outputs was located on any readable surface. The trust centre describes how data is held rather than for how long, covering AES-256 encryption at rest, TLS 1.2 or higher in transit, single-tenant deployment with a dedicated instance per customer, role-based access control and redundant backups, and states that the customer owns its data. None of that is a retention position. No default period, no configurable window, no deletion route and no certificate of destruction appears anywhere readable, and nothing addresses what happens to a fund's side letters at the end of a matter or of the subscription. The documents that would carry it are listed and gated: a Data Processing Agreement, a Privacy Policy and a GDPR Overview, each resolving to a mailto link. Searched the home page, the ProVision overview, the about page, the support FAQ and the trust centre on 4 September 2026.

Source: Operator VerifiedAs of Sep 4, 2026

Ethical Walls and Matter Segregation

Does retrieval respect the firm’s ethical walls, or can the model read across them?

Own model, documented

The product maintains its own permission model, documented, requiring the firm to keep it aligned.

A separation model is documented rather than asserted, and the vendor makes it a positioning claim. The trust centre states that most legal technology providers operate shared multi-tenant environments and that ProVision is different, being deployed within a secure single-tenant AWS environment logically segregated per customer, with each customer's documents and extracted data held in a dedicated instance and never commingled with other clients' data. It lists no shared document storage across customers as a distinct control, alongside granular role-based access control limiting access to authorised users. That answers separation at the tenant level, which is the level a law firm buyer requires of a hosted platform. One limb below it is not addressed: nothing describes segregation between matters, funds or clients inside a single firm's instance, which is a live question on a product whose value is comparing provisions across investors and across matters, and no ethical wall or conflicts mechanism is described.

Source: Vendor Publishednever commingled with other clients' dataAs of Sep 4, 2026Evidence

Third Party Request and Subpoena Notice

If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?

Not addressed

No located term or policy addresses third party requests for customer data.

Nothing located addresses compelled disclosure or customer notice in either direction. The evidence home for this signal is the confidentiality section of a master agreement or the law enforcement section of a privacy policy, and neither is readable: the Terms of Service and the Data Processing Agreement are listed on the trust centre behind mailto links, and the Privacy Policy appears both behind the same gate and at a public footer URL that was not opened in this pass. No transparency report, government request statement or disclosure practice appears on the home page, the ProVision overview, the about page, the support FAQ or the trust centre body, all read on 4 September 2026. The nearest adjacent statement runs to access rather than disclosure, with the trust centre asserting a zero-knowledge framework under which ILS cannot access customer provisions or investor data; that claim, if it holds, would bear on what the company could produce under compulsion, but it is not a notice commitment and nothing published connects the two.

Source: Operator VerifiedAs of Sep 4, 2026
Accuracy and Authority

Primary Law Corpus Provenance

Where does the law in this product come from, and does the vendor have the right to use it?

Not addressed

No located public material identifies the corpus behind the product’s answers.

The inputs are named and they are the customer's own documents rather than any licensed corpus. Published material describes the platform working from uploaded side letters, limited partnership agreements, related documents and existing master side letters, with provisions structured from them while the original legal text is preserved. There is no external body of content: the product does not retrieve primary law, published precedent, a market-standard clause bank or any third-party dataset, and none is named on any surface. The one place a corpus question could arise is the comparison and merge feature, which surfaces provision information across clients and matters, but that operates on the firm's own material rather than on licensed content. There is accordingly no licensing question of the kind this signal was written for and no jurisdictional coverage statement to record. Searched the home page, the ProVision overview, the about page, the support FAQ and the trust centre on 4 September 2026.

Source: Vendor PublishedAs of Sep 4, 2026

Good Law Verification

Does the product tell you when the authority it just cited has been overruled?

Not addressed

No located public material addresses whether authority is checked for subsequent history.

Nothing on any located surface addresses checking authority for subsequent history, and the product does not retrieve or present primary law at all. ProVision reads investor side letters and fund documents and produces structured provisions, master side letter views and MFN election forms; no case, statute or regulation is surfaced to a user at any point in the published workflow. The question does not bite on this product class and the value records the honest absence rather than a shortcoming. Searched the home page, the ProVision overview, the about page, the support FAQ and the trust centre on 4 September 2026.

Source: Operator VerifiedAs of Sep 4, 2026

Refusal and Uncertainty Behaviour

What does the product do when the answer is not in the corpus?

Not addressed

No located public material addresses what the product does when it cannot ground an answer.

An assurance is published and no behaviour is described. The support FAQ states that the platform provides fail-safes or guardrails to retain full control over outputs and prevent hallucinations, which addresses the topic in the abstract, and two design controls sit alongside it: extraction preserves the original legal text, and comparison is said to operate without breaking traceability, so a reviewer can return to source. None of that describes what the system does when it cannot ground an output. No abstention state, no no-answer condition, no confidence or certainty score against an extracted provision, and no described handling of the ordinary failure conditions for this product class is published, and those conditions are specific and foreseeable here: an ambiguous or hand-amended side letter, a scanned execution copy, a provision that spans clauses, or two investor terms that conflict. A claim to prevent hallucinations is an assertion about the outcome rather than a description of the behaviour, and is recorded in this summary rather than credited as one.

Source: Vendor PublishedAs of Sep 4, 2026

Fabricated Citation Record

Does a public court record exist involving output from this product?

None located

No court order, opinion or disciplinary record naming this product has been located as of the date shown. This is a statement about the public record, not a finding about the product.

The AI Hallucination Cases database maintained by Damien Charlotin was searched on 4 September 2026 on the product name ProVision and on the company names Intelligent Legal Solutions and ILS. No court order, opinion or disciplinary record naming the product or the company was located. This records the state of the public record on that date and is not a finding about the product. The signal also sits at an angle to this product class, since ProVision structures negotiated investor terms rather than generating legal citations, so a fabricated citation is not the failure mode it would ordinarily produce.

Source: Operator VerifiedAs of Sep 4, 2026Evidence
Professional Responsibility

Bar Guidance Alignment

Has the vendor engaged in public with the ethics opinions its buyers are bound by?

Not addressed

No located public material engages with bar or ethics guidance.

No bar authority, regulator, conduct rule or ethics opinion is named on any located surface. Nothing on the home page, the ProVision overview, the about page, the support FAQ or the trust centre engages professional regulation, and no jurisdiction-specific guidance is mapped. The absence is more consequential than it would be on a domestic product because the company reports operating across the United States, United Kingdom, European Union, Hong Kong, Singapore and the Middle East, which carry materially different regimes for the use of technology on client matter material and for cross-border handling of it. The nearest published language addresses professional judgement rather than professional rules, with the about page stating that the technology assists rather than second-guesses professional judgement. The Terms of Service, which is where a jurisdiction or conduct provision would ordinarily sit, is available only by emailing the company.

Source: Operator VerifiedAs of Sep 4, 2026

Billing and Fee Posture

Does the vendor address what happens to the bill when the work takes an hour instead of six?

Savings claims only

Public materials claim time savings without addressing billing or disclosure.

Recovery and efficiency claims are published and nothing addresses the bill itself. The support and trust centre FAQ frames the problem in explicitly billing terms, stating that manual transfer of side letters into Excel or Word compendia and manual MFN tracking results in hours billed that are not recovered by the firm, inaccurate work product and slower delivery, and a further collapsed entry is headed What is the ROI. So the pitch is that the firm captures fees it currently loses. What is absent is everything this signal asks for: no per-matter record of AI-assisted work is described as available, no guidance on fee or disclosure treatment is published, and nothing states whether a provision extracted or structured by the model is identified as such to the firm or to the fund client whose matter it is. The direction is worth recording, since it is the second record in this pull to run the opposite way to the compression this signal was written to catch: the stated effect is recovering unrecovered hours rather than reducing them.

Source: Vendor PublishedAs of Sep 4, 2026

Outside Counsel Guideline Readiness

Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?

On request only

The material exists behind a sales conversation or an executed agreement.

The artifacts a firm would need exist, are named individually, and every one of them is behind an email request. The trust centre's Resource Vault lists a Data Security White Paper described as a technical deep dive into architecture, encryption and the zero-knowledge framework; a GDPR Overview stated to cover where data is hosted and which subprocessors are involved; an independently issued SOC 2 Type II report; a Privacy Policy; a Data Processing Agreement setting out the company's role as processor; and an Incident Response Overview. Each View link resolves to a mailto address at the company's information security team, and the portal does not state whether the request fulfils without a sales conversation. That is precisely this value: the material exists and is reachable only by asking. It does not reach the rung above, because no subprocessor or model provider is named anywhere in the clear, so a firm cannot answer a client's AI clause from published material alone. The only external link in the vault points to the California Attorney General's own CCPA page rather than to any ILS document.

Source: Vendor PublishedAs of Sep 4, 2026

Court Disclosure Support

If a judge’s standing order requires an AI disclosure, can the product produce one?

Partial record

Some elements of the record are available, short of a document level export.

An audit record is published as a product feature and nothing states that it distinguishes model work from human work. The home page commits to keeping a clear, complete record of every action, approval and document version to support compliance, and the overview describes comparison and structuring operating without breaking traceability, with the original legal text preserved so an entry can be traced to its source clause. That is a real record of what happened to a document and who approved it, which is more than most records in this corpus publish. What is absent is the AI-specific half. Nothing says the record identifies which provisions were extracted by the model rather than entered by a person, no model or version is attributed to an output, and nothing describes exporting that history for a client audit, a regulator or a court. Exports that are described are of work product, being MFN election forms in PDF and Word with built-in anonymisation, rather than of the audit trail.

Source: Vendor PublishedAs of Sep 4, 2026
Contact

Correct a record, or ask how something was graded

Every grade and every signal on this index is drawn from public sources and dated. If a record is wrong, out of date, or missing an artifact the index did not locate, send the source and it will be reviewed and the record redated. Vendors are welcome to submit documentation. Nothing on this index is for sale, including a listing, a placement, or a grade.

AI Legal Index

The AI Legal Index is an independent index that tracks changes to AI vendors in legal. It holds 61 vendors across 9 categories, each graded on the same 15 capability axes and recorded against 12 legal signals, from public sources. No vendor pays for inclusion, placement, or rating.

Index Status
Last index update
September 5, 2026
The AI Legal Index is an editorial reference. It is not a regulatory body, not a law firm, and nothing published here is legal advice or a recommendation to retain or avoid a vendor. Records are verified against published sources, bar guidance and public court records. Where a record reads not addressed, the material was not located in public sources on the date shown. See the Methodology page for evaluation standards and limitations.
© 2026 AI Legal Index
3801 N Capital of Texas Hwy, Ste E240 · Austin, TX 78746