Jimini AI vs Ordalie: how they compare in 2026

The CGU defines the thing being sold as the "Système d'IA" and the Site as an interface for interacting with it, which is as literal as this axis gets. Analysis, drafting and research are all model outputs, and the agentic workflow layer plans and orchestrates multi-step tasks. The home page describes a multi-model architecture in which Jimini selects, orchestrates and continuously optimises third-party models per task and builds legal reasoning, source handling and workflows on top. Remove the models and there is no document system or workflow engine left underneath. Checked 4 September 2026.

Remove the models and there is nothing left to sell, which is the A band. The agreement itself settles it rather than the marketing: Terms of Sale clause 2.1 defines the subscription as three things and all three are model-run, an AI assistant that answers legal questions and analyses documents, a search engine described as using AI algorithms to understand the links between legal documents, and intelligent document generation producing contracts, memos and analyses. There is no fourth, non-AI service in the definition; the only other item is training and support in the use of those three. The corpus underneath is real and large, and the tempting counter-argument is that a searchable database of 102 codes and case law would still have value with the models switched off. It does not carry, for the vendor's own reason: that database already exists free at Legifrance, and Ordalie positions expressly against it, naming manual Legifrance research as the thing it replaces. What the buyer pays for is the synthesis and the citation, not the texts. Pricing follows the same logic and is stated in model terms, the free and paid tiers being separated by which legal AI models are available, basic against advanced, and by query volume rather than by any non-AI feature. Verified 12 September 2026.

Grounding is real and documented rather than asserted. The research product is described as returning answers anchored in recognised sources of law, Legal Data Hunter is named in the CGU subprocessor table as the legal research provider located in the European Union, and the privacy policy describes the retrieval mechanism directly: uploaded documents are indexed and split into paragraphs to support search and retrieval. What keeps this off the top band is the absence of a measured figure a reader can interrogate. The home page carries a precision claim in a hero counter, presented with no methodology, no test set and no date; a number in that form is a claim shape rather than a published accuracy figure, so it would not lift this axis even if it were read. No failure modes are named anywhere. Separately, and as a limit on this record rather than on the vendor, the counter's numerals render through an animated component that did not resolve on fetch. Checked home page, product pages, workflows page, CGU and privacy policy on 4 September 2026.

Grounding is real, documented and openable by the reader, with a described retrieval method and no published measurement, which is the B band precisely. The grounding limb is met more fully than most records in this corpus reach: every answer displays its sources alongside it, citations carry direct links to the original text, references are given to the article, judgment number or page, the excerpts actually relied on are shown, and where possible the link resolves to Legifrance so the reader lands on the official version rather than a vendor rendering. The retrieval method is described rather than asserted, the documentation stating that the assistant selects the relevant sources itself, shows the user which ones it selected, then sorts and synthesises them into an answer with matching citations, with legal references detected and formatted in real time as the answer is written. The user controls the retrieval perimeter by switching source families and jurisdictions on and off. What is absent is everything A adds. No accuracy figure, error rate, test set or evaluation of any kind is published, so nothing here is testable by an outsider. Citation status is not checked, which is graded on the citator signal. And no statement was located describing what the system does when it finds no support. R40 governs the floor and the D limb does not fire: the no-hallucination claim on the marketing sits alongside architecture that is documented in detail, so it is not a bare claim. Verified 12 September 2026.

The CGU preamble is emphatic and sits in bold at the top of the agreement: outputs are not legal advice, are not final, must be validated by a legal professional such as an avocat, and the user assumes entire responsibility for their use. Against that, the product markets agentic systems that plan, orchestrate several steps and carry legal tasks through from end to end, with more than 200 workflows designed and validated with in-house lawyers. Those two statements are in tension and nothing reconciles them. What the tension reveals is the limb that is missing rather than a reason to downgrade for its own sake: no threshold is published at which an agent stops and hands back, no review surface is described inside the workflow, and nothing states what happens after an agent is wrong. The contractual commitment to human validation is real and the control structure around it is not published.

A written commitment that the product works alongside a supervising lawyer, with real review surfaces, short of any threshold at which it proceeds alone. The review surfaces are concrete and are the strongest part of this row: every answer carries clickable source dots the user opens to confirm the detail, the editor ships track changes where every modification is traceable, commentable and reversible, and the Ledger documentation tells the user in terms that its filters are there to help review and not to replace checking the cited source. The route back to human judgement is contractual, Terms of Use 5.2 stating that Ordalie is not authorised to provide legal services and that the texts, positions and analyses it presents do not constitute legal recommendations or professional legal advice, and encouraging the user to seek the opinion of a duly qualified professional. Human oversight is also one of the five named dimensions of the CNB alignment claim. R37 rule 2 governs what holds it off A. The same estate markets workflows that return a finished deliverable ready to export as PDF or share with your client, and press material describes automating up to eighty per cent of repetitive tasks, which cannot sit comfortably alongside a review-every-answer posture. The conflict is not itself the downgrade; what it reveals is that no threshold, confidence boundary or class of work is published at which the system runs without a reviewer, and no statement addresses what happens after it is wrong. That is the limb the B band names. Verified 12 September 2026.

Named deployment evidence with substance, short of measurement. A case study on the vendor's own blog dated 26 June 2025 names Hoche Avocats, describes a three-month trial followed by adoption across corporate, tax, employment, technology and litigation departments, and sets out how the rollout was co-managed by partners and associates and how prompts were built from the firm's own agreement frameworks and M&A audit reports. Attributed quotes name Louis Oudot de Dainville, partner at Gide Loyrette Nouel, and Jacques Perotto, partner at Alerion. The Paris Bar partnership is dated May 2025 with its terms stated, three months' free access for firms of one to twenty lawyers, and Pierre Hoffman is quoted as Batonnier. A user count of more than 6,000 avocats and juristes is published. What is absent is a documented outcome: the Hoche study reports no measured result, and the hours-saved-per-lawyer figure on the home page sits in a hero counter with no method, no cohort and no date behind it, which is a claim rather than a measurement and would not carry this axis to the top band in any event. As a limit on this record rather than on the vendor, that counter's numerals render through an animated component that did not resolve on fetch on 4 September 2026.

Real deployment evidence with substance on both limbs separately and neither joined to the other, which is the B band in its exact terms. Customers are named: a logo strip headed as legal teams who use Ordalie daily carries Mondial Relay, Radio France, Ifop, Aramis Group, the Congres des Notaires de France, Editions Tissot, Carlini Avocats and Gouache Avocats, a mix of corporate legal departments, an institution and two law firms. Three testimonials are attributed to a named individual and organisation, including the legal department of SNCF Connect and Tech and the Congres des Notaires de France, which under R122(2) is named customer evidence rather than outcome evidence and is counted here as attribution, not as a result. Figures are published separately: eight hours saved per week on average, a case timeline in three minutes against a stated manual baseline of three hours, and an hour instead of a full day. The two never meet. Not one figure is attached to a named customer, no date appears against any deployment or any figure on the surfaces read, and no method is published for the eight-hour average, so a reader cannot tell what population it was drawn from or over what period. That is precisely what the B band describes, a named customer without figures and figures without the named customer. The case studies and news estate were not opened; under R25 they corroborate rather than carry a row standing on the strip itself, and they are the artifact that would move this row if they carry dates and a stated basis. Verified 12 September 2026.

Every limb is met and read before signing. Training is prohibited in the CGU itself and the prohibition extends to suppliers. Segregation is documented at both levels the buyer segments need: data is logically partitioned by user and by organisation, personal and enterprise libraries are stored in isolation, and access runs through Auth0 by Okta on a least-privilege basis with access logged and audited. Professional secrecy is addressed in a dedicated clause rather than by implication, citing article 66-5 of the law of 31 December 1971 and article 2.3 of the Règlement Intérieur National, acknowledging the absolute and public-order character of the obligation, and stating that no data in submitted documents is accessible to third parties or used for any purpose other than providing the service. Retention and deletion are published as a per-category table with a certificate of destruction available on request. The position on model providers is the strongest part: each inference provider is named with its location and the specific basis of its zero-retention engagement. The middle band's two named gaps, silence on segregation between users and silence on what the model provider may retain, are both false of this record, which is what decides it.

Four of the five A limbs are met and met contractually, and the fifth fails, so the conjunctive band gives B. This is the closest call on the record and the note names the whole of it. Training: Terms of Sale 9.1 states flatly that client data is not used to train models, which is the agreement rather than a policy page. Segregation: case-based spaces, projects gathering a matter's material, role-based permissions, personal connections kept private from organisation connections, personal tags, and a statement that a user's own sources are never accessible to other users. Privilege addressed directly, which R33 makes decisive and which most records fail: legal privilege is named as the governing constraint on the lawyer-facing surface and carried as its own security commitment, and clauses 7.1 and 9.3 give it a mechanism rather than a slogan, an irrevocable undertaking binding Ordalie and its employees not to use or reveal the content of documents transmitted, with liability expressly acknowledged for breach, and a single narrow exception for salaried developers correcting a processing error, limited to files that errored, refusable by the customer, with correction copies irretrievably destroyed twenty-four hours after the error. Third-party model providers: made explicit, the inference providers being contractually required to process in France, not to reuse client data for training and not to retain prompts, outputs or associated metadata. The limb that fails is retention and deletion stated. Deletion is stated well; retention is not, and worse, it is stated inconsistently, the pricing matrix offering unlimited history on the paid tiers while the documentation says conversations are kept for thirty days, and the agreement giving no period at all for uploaded documents. Verified 12 September 2026.

The CGU preamble states in bold that outputs are not legal advice and must be validated by a legal professional, that Odysai is not a law firm and provides no legal advice, and then names the governing statute: under the law of 31 December 1971 as amended, Odysai, its officers and its employees do not carry on legal consultation, the drafting of private deeds, or representation. Naming the French UPL statute directly is rare. The audience is limited expressly to users acting in a professional capacity, and consumer mediation under articles L.611-1 of the Code de la consommation is disapplied on that basis, so there is no consumer surface to disclose on. The jurisdiction is stated throughout as continental and French law. The vendor's own help centre article on déontologie engages the CNB's recommendations and the EU AI Act compliance timetable. Competence and supervision are carried by the validation requirement rather than by a separate competence statement, which is the one place a reader must infer.

A real published position on advice against tooling, reaching further into the supervision dimension than most records at this grade, short of the treatment A asks for. The advice line is drawn in the agreement rather than in a footer, Terms of Use 5.2 stating that Ordalie is not authorised to provide legal services, that the texts, positions and analyses presented, suggested or found are not legal recommendations or professional legal advice, and directing the reader to a duly qualified professional. That is specific about what is disclaimed and it is readable before signing, so C does not fire. The supervision dimension is touched rather than left out, which is unusual: the vendor states that the product is designed to align with the French national bar council's guidance on AI use by lawyers and enumerates human oversight and traceability among the dimensions, which engages the duty rather than merely disclaiming advice. What holds it off A is who may use it and where. Three professional audiences are addressed by name, lawyers, in-house counsel and notaries, but the free plan is open self-serve to anyone with an email address and nothing published states that the product is for professionals only or addresses use by a non-lawyer, the nearest thing being a contractual bar in clause 3.1 on registering under a false professional status, which polices accuracy of a declaration rather than eligibility. No jurisdictional limit on reliance is stated, on a product answering across five named jurisdictions, and nothing describes how a supervising lawyer discharges a competence duty over its output. Verified 12 September 2026.

This is the weakest disclosure on an otherwise strong record. There is no AI-specific governance framework published: no ISO 42001, no responsible AI page, no named owner accountable for model governance inside the company, no pre-release testing regime described, and nothing at all published about uneven output across matter types or populations. The vendor's help centre article on déontologie engages the EU AI Act timetable and states that technical documentation, the DPA and the security policy are available on request, which places the governing artifacts behind a request rather than in public. ISO 27001:2022 is real and certified but is a security management standard and does not answer this axis. The DPO, Auriane Roussel, is named for data protection, which is a different accountability from AI governance. Checked home page, CGU, privacy policy, trust centre and help centre on 4 September 2026.

A position exists and no mechanism, testing regime or accountable owner is published, which is C. What is located is an alignment claim rather than a governance programme: the vendor states that the product is designed to align with the CNB's requirements for AI use by lawyers and lists transparency, confidentiality, human oversight, data protection and traceability as the dimensions engaged. That claim is graded on the bar guidance signal, where it earns the named-guidance value, and it is deliberately not credited a second time here, because working one fact across two rows is the error the ground rules warn about most. Read as governance it is an assertion of compliance with an external framework, unaudited, with no mapping document, no published assessment against the seven requirements and nothing a buyer could inspect. R36 is the calibration point and it cuts against a B: ISO 42001 earned a B because it is independently audited, and there is no ISO 42001 here. SOC 2 Type II and ISO 27001 are security attestations and are graded on the certifications row. Nobody inside the vendor is named as accountable for AI, nothing is published about what is evaluated before a model or a workflow ships, and no incident or failure has been disclosed. Bias is addressed nowhere, and the shape it takes on this product is worth naming: answers are assembled from source families the user switches on and off across five jurisdictions of very different depth, so output quality is structurally uneven across practice areas and jurisdictions, and nothing published examines that. Verified 12 September 2026.

The full set is published and specific enough to hold the vendor to. Retention is a per-category table: document library and AI prompts and responses for the life of the account plus 30 days, account data one year after last connection, billing data ten years, Word and Outlook add-in document and email content not stored at all, connector data processed in real time with zero storage. Deletion is specific, with account deletion executed within the day, aggregated usage data removed within the month, and a certificate of destruction issued on request. Access control names Auth0 by Okta, least privilege, logged and audited access, AES-256 at rest and TLS 1.3 in transit, logical partitioning by user and organisation, and vulnerability testing every six months. Subprocessors are fully named in a table with role, location and zero-retention status, with 30 days' advance notice of any addition or replacement, a right to object, and a right to terminate without penalty if an objection is not resolved. Incident practice is stated on both sides: notification to the customer within 24 working hours of qualifying a breach and to the CNIL within 72 hours under article 33.

Four of the five limbs are published and specific and the fifth is both vague and self-contradictory, which is B. Access control: AES-256 at rest and HTTPS/TLS in transit, account isolation, access to content limited to authorised persons and logged where needed for security, operation or support, role-based permissions, and SSO/OIDC with access governance on the top tier. Subprocessors: a genuine named list in the privacy policy at article 3.4, Scaleway for hosting, AI inference providers including Azure France, and Stripe for payment, each with its function stated, updated 6 June 2026. Incident practice: Terms of Sale 9.4 commits to notifying the customer without undue delay of a personal data breach so the customer can meet its own obligations. Deletion: unusually precise, content erased immediately and irreversibly from production environments when a user deletes it or the applicable period ends, with residual copies acknowledged as remaining in encrypted backups until overwritten, which is a more candid account of backup reality than this corpus usually sees. Retention is the gap and it is a real one. No period is stated for uploaded documents anywhere in either agreement, the contractual formula being the period applicable to the relevant workspace or organisation without saying what that period is or who sets it. The documentation states that conversations are kept for thirty days while the pricing matrix sells unlimited history on both paid tiers, and nothing reconciles them. The retention-side commitment that is specific runs to the inference providers rather than to Ordalie: they are contractually barred from retaining prompts, outputs or associated metadata. Verified 12 September 2026.

This is the sharpest gap in the record and it sits against an otherwise exceptional set of published commitments. The CGU's limitation of liability clause is an exclusion and nothing more: the company is not liable for damage arising from unavailability caused by factors outside it, from use of outputs without prior validation by a qualified legal professional, from user breach, or from external attack. The intellectual property clause adds that the company cannot be held responsible for outputs or their use, expressly including error, approximation, or content a third party may claim. There is no indemnity running to the customer, no liability cap expressed as a figure or a multiple, no warranty of any kind on the service or its output, and no insurance position was located on the home page, CGU, privacy policy, legal notices or trust centre on 4 September 2026. What is published is a clear allocation that the user bears the loss, and a buyer can read it before signing, which is why this sits at C rather than at the floor. The exposure the product creates is erroneous legal output and the clause disclaims precisely that.

A real published position on liability that goes beyond disclaiming, short of anything that reaches the exposure the product actually creates. C does not fire, and the word that decides it is C's own: liability is not addressed only through a limitation clause. Two provisions go further. Terms of Sale 7.2 is an undertaking to carry professional liability insurance and any policy covering the risks of its activity, which is a published insurance position and is rare in this corpus. Clauses 7.1 and 9.3 close the confidentiality undertaking with an express acknowledgement that Ordalie may be held liable for breach of it, which is a vendor accepting liability by name rather than excluding it. The limitation itself is comparatively narrow as drafted, Terms of Use 5.4 limiting liability to direct and proven damages rather than excluding whole categories of loss. What keeps it well short of A is that none of it is quantified or invocable and none of it touches accuracy. No indemnity of any kind was located. No cap figure, no insurer, no cover limit and no statement that the customer benefits from the policy. No warranty of output, and the opposite: Terms of Use 5.1 supplies the site on a best-efforts basis with no obligation of result and tells the user they proceed at their own risk, and Terms of Sale 8.2 makes the user solely responsible for their use and disclaims responsibility if the service does not meet their needs. The availability target of 99.95 per cent is expressly stated not to be an obligation of result and not to found any liability action, which sits against the pricing page selling priority support with an SLA on the top tier. So a buyer relying on a wrong answer has recourse to nothing published. Verified 12 September 2026.

Integrations are into the systems legal work actually lives in, and the depth is documented rather than listed. Jimini pour Word and Jimini pour Outlook each have their own product page and are included in the subscription, with functions named per surface: drafting, analysis, correction, comparison and translation in Word; drafting, thread summarisation and text improvement in Outlook. Connectors run to Google Drive and SharePoint. The privacy policy then documents each integration's data flow to a level most vendors never publish: for the add-ins, document and email content including attachments is processed and discarded rather than stored, with only the prompt and the generated response retained; for connectors, only files the user selects are accessed, processing is real time with no storage beyond immediate use, access is revoked immediately on disconnection or account deletion, and Google connector handling is stated to follow the Google API Services User Data Policy. What syncs, in which direction, and what the firm must enable are all answerable from published material. No legal-specific document management system such as iManage or NetDocuments is named.

Documented, verifiable integrations into the systems legal work already lives in, with direction and configuration described, which is every limb of the A band. Eighteen connectors are named and each carries its own documentation page: SharePoint, Workspaces, OneDrive, Outlook, Gmail, IMAP, Google Drive, Dropbox, Box, Nextcloud, Tomorro, Eudonet, Odoo, Nuxeo, iManage, Secib Neo, Microsoft Teams and Managed APIs, with Agestia named alongside. Two of those are the systems this axis exists to test: iManage, the document management system a large part of the profession runs on, and Secib Neo, a French law-firm practice management platform. Direction is stated rather than implied, and stated per connector: iManage is brought in read-only through OAuth using credentials generated in iManage Control Center, Dropbox and Box can be granted write access so Ordalie saves produced documents back and creates folders, and the general rule is published that a read connection browses and previews while a write connection returns output to the source. What a firm must configure is published to the field: tenant, site, client ID and secret or a certificate for organisation-wide SharePoint, WebDAV URL and credentials for Nextcloud, instance URL, optional database name and API key for Odoo, and a documented choice for iManage between one shared connection and each member signing in individually. Personal against organisation scoping is documented throughout, as is revocability. Two limits are recorded and neither is an A limb: there is no public developer API, the managed APIs being described as for AI-assisted work rather than a general developer console, and two-way synchronisation is routed to the separate Vault module. Verified 12 September 2026.

Residency is stated flatly — all customer data is stored in France — and then evidenced rather than asserted. The platform at app.jimini.ai is hosted by Scaleway and S3NS, both named with Scaleway's registered address given and both stated to hold servers physically located in France, Paris region. Processing location is published as distinct from storage location and is named per inference provider: Mistral AI in France, OpenAI's Europe deployment in the European Union, and Anthropic Claude models through Google Cloud Vertex AI in Belgium at europe-west1. Tenancy is stated separately as logical partitioning by user and by organisation. The marketing site itself is documented as setting no audience-measurement cookie and storing nothing on the visitor's device. The band's limb on what changes between tiers has nothing to bite on, because Jimini publishes a single hosted deployment in a single country rather than a set of options, and the limbs that do apply are answered to the cloud region.

Regions, tier differences and the processing-against-storage distinction are all published, which is the full A band and makes this the strongest residency disclosure located in this pull. The region is named and then named again at the level below: hosting is in France, and the host is identified as Scaleway with its corporate registration and its registered office given in both agreements, with the privacy policy going as far as the specific site, Paris 2. What changes between tiers is published on the pricing matrix rather than left to a sales conversation: on-premise deployment is marked absent on the free and PRO tiers and present on MAX, alongside SSO/OIDC and access governance, and the enterprise route is described as adding on-premise or private hosting and custom connectors. The limb almost every record in this corpus misses is the one this vendor answers most deliberately. Where data is stored and where it is processed are addressed separately and by different mechanisms: storage sits with Scaleway in France, while inference is performed by third-party providers, including Azure France, that are contractually required to process the data in France. So a buyer is told not merely that the service is French-hosted but that the model call does not leave the jurisdiction either, and is told which instrument makes that true. Recorded and not credited, because it belongs to other rows: the same clause bars those providers from retaining prompts and outputs, which is stewardship, and identifies them, which is supply chain. One honest limit on the record: no region other than France is offered, so this is a single-region product, and the disclosure is complete rather than the choice being wide. Verified 12 September 2026.

Better evidenced than most records in this band and short of the top on two limbs. A trust centre is live at trust.jimini.ai, publicly reachable with no sales call, no email capture and no NDA click-through, listing the entity as ODYSAI, showing ISO 27001:2022, enumerating 114 controls grouped under business operations, governance, asset management and risk management, and reporting a last refresh a day before the check. The DPA goes further than the portal by publishing the certificate number IC-IS-2506288 and its validity to 17 June 2028, which lets a buyer verify the certificate independently. Two limbs are missing: no certification body or auditor is named on any surface read, and no attestation report was located as obtainable — the portal lists a single document without identifying it in what rendered. This is the R16 band gap appearing from the opposite direction, a record with a public portal, a certificate number and an expiry date but no named auditor and no reachable report. Checked 4 September 2026.

Certification is real, named and stated in the vendor's own voice, with no route to the evidence, which is the B band. Two standards are claimed, SOC 2 Type II and ISO 27001, and they appear on the lawyer-facing page, in the answer to a direct FAQ question about client data protection, in a certifications panel and in the site footer. C does not fire, for two reasons that are worth separating. These are not unexplained badges: the claim is made in prose, in response to a question a buyer would actually ask, and it is accompanied by an infrastructure panel giving a last audit of January 2026 and a commitment to annual security audits, so a date of sorts is published. And R16 does not bite the way it did on comparable records: the subject of the sentence is Ordalie, not its host, and the vendor does not reach for Scaleway's certifications to stand in for its own. What is missing is the whole evidentiary apparatus that A requires. No auditor is named for either standard. No report period or observation window is given, no certificate or report number, and no scope statement saying which systems or trust services criteria the SOC 2 covers. There is no trust centre of any kind, and no route to a report at all, not even a sales-gated one: the site's own Security link in the footer resolves to the privacy policy rather than to a security page, so the page inventory contains no security surface. R5 is therefore not reached, because there is no access flow to grade. Nothing is published on penetration testing, vulnerability remediation or an incident response function. Verified 12 September 2026.

The most complete supply chain disclosure this pull has seen. The CGU publishes a subprocessor table naming every provider with access to customer data, its role, its location and the specific basis of its zero-retention engagement: Mistral AI SAS for inference in France on documented activation, OpenAI's Europe deployment in the EU under a signed contractual amendment, and Google Cloud Vertex AI carrying Anthropic Claude models in Belgium europe-west1 under written supplier approval, alongside Scaleway and S3NS for hosting, Gradium SAS for voice transcription, DeepL for translation, Legal Data Hunter for legal research and Cloudflare for transit only with no storage. The privacy policy repeats the inference provider list independently. Change notification is committed at 30 days with a right to object and to terminate without penalty. One limb is only partly satisfied and is named here rather than glossed: Anthropic Claude is identified as a model family, but the specific Mistral and OpenAI models are not named, and the home page describes continuous selection and orchestration of the best models per task, so which model handles a given request is not disclosed. Adjudicated to B on 4 September 2026 under R34: provider identification and model naming are separate limbs of the A band, and the second is not met. The provider disclosure remains the strongest in the pull and the grade reflects the missing limb only.

A provider is named and no model is, with no change notification, which is the first limb of the B band exactly. What is disclosed is more than most: privacy policy article 3.4 identifies AI inference providers, including Azure France, as a category of recipient, states what they are contractually required to do, and Terms of Sale 9.1 repeats the naming, so a buyer learns that inference runs through Microsoft's French Azure estate and not merely that advanced models are used. Where the models run is stated, in France, contractually. R34 governs the rest and holds this at B: the A band lists the models being named and their providers being identified as separate limbs, and no model or version is named anywhere. The vendor's tier language, basic against advanced legal AI models, is a commercial label rather than a disclosure. No commitment to notify customers when the model or provider changes was located; the fifteen-day notice in Terms of Sale 2.3 attaches to deletion or major modification of a service, which is a different object, and the subprocessor list is framed as accurate as of the date of the policy update rather than as maintained. The genuine tension on this row is recorded rather than resolved: the company's own positioning, and press coverage repeating it, is that Ordalie develops its own models, while its published documents name third-party inference providers. Both can be true at once, but nothing published reconciles them, so a reader cannot tell whether a given answer came from a model Ordalie built or one it rented. Verified 12 September 2026.

Two named plans are published with their feature splits: an individual plan carrying all tools, the Microsoft extensions, AI workflows and data confidentiality, reachable through a self-serve signup with a two-week free trial and no sales contact; and an organisation plan adding collaboration features, training and onboarding, a usage dashboard and centralised billing, routed to the sales team. The CGU establishes the charging terms directly — self-serve rates are those displayed on the site at subscription, expressed in euros excluding tax, billed monthly or annually at the subscriber's choice, with 30 days' notice of any price change and a right to cancel before it takes effect. The individual figure itself was not read: the pricing page renders its numerals through the same animated counter component that obscured the home page statistics, and a search of the indexed page did not recover them. The CGU confirms figures are displayed at the point of subscription, so this is a limit on the retrieval rather than a gap in the publication, and it is recorded as such. Real pricing published for part of the range with the enterprise tier withheld is the middle band exactly.

A buyer can learn what this costs without speaking to anyone, which is the A band, and the surrounding commercial mechanics are published to a level few records reach. Rates: a free plan at zero, PRO at 75 euros excluding VAT per user per month, MAX at 99 euros on the same basis, with annual billing at a stated ten per cent discount, giving 67.50 for PRO. The unit is named explicitly and repeatedly, per user and per month, and the VAT treatment is stated. The structure is set out in a twenty-six row comparison across the three tiers covering models, query volume, history, source limits, uploads, exports, collaboration, integrations, governance and support, with an explicit legend for included, limited and not included. What implementation adds is answered for the self-serve range: purchase is by card or SEPA through a Stripe portal, seats are added or removed by an administrator with immediate prorated effect, plan changes prorate in both directions, monthly carries no commitment and cancels in a few clicks with no notice period, annual is billed once for twelve months, and invoices are consolidated and self-downloadable. Two things are recorded and neither displaces the grade. The enterprise route beyond MAX is quote-based, covering Vault, on-premise or private hosting, custom connectors and pilots, which is the ordinary withholding at the top of a published range. And the same page contradicts itself about what the top tier includes, the matrix ticking the Vault module and on-premise deployment for MAX while the FAQ places both beyond MAX in the enterprise offer. The Terms of Sale add a third conflict, stating prices include VAT at twenty per cent where the pricing page says all prices exclude it. Verified 12 September 2026.

Segment coverage is described with real substance. The pricing page addresses independent avocats and organisations, splitting the latter into law firms and corporate legal departments, and the bar partnerships size the segment precisely, with the Paris agreement covering firms of one to twenty lawyers. The customer set spans both sides, naming law firms such as Gide Loyrette Nouel, Capstan, Hoche Avocats, FTPA and Franklin alongside corporates including CMA CGM, Petit Forestier, Léon Grosse and In Extenso. Practice coverage is evidenced through the agentic workflow library, which is filterable by affaires, civil, immobilier, pénal and public. The jurisdictional boundary is stated plainly and repeatedly as continental and French law. What is left open is the far edge: no statement says which practice areas are not supported, government use is not addressed, and the five workflow filters are a library taxonomy rather than a declared scope of support.

Who this serves and what it covers are described with real substance, and the boundaries are left open or stated inconsistently, which is B. The audiences are three and each has a page of its own rather than a line in a list: lawyers in private practice, in-house counsel, and French notaries, the last being a genuinely distinct professional segment this corpus has not seen addressed before, supported by named notarial customers and workflows built for notaries. Collaboration is described at firm level, with shared case files, role-based permissions and a private space per lawyer. Practice coverage is enumerated rather than claimed: the source documentation lists the 102 French codes by category, names the courts whose decisions are carried, identifies BOFiP for tax, and describes how a practitioner narrows the perimeter, with a worked example of a business-law specialist switching off the public health and urban planning codes. Real limits are published, which is more than most: jurisdictions are selectable rather than assumed, uncovered countries are requestable, and the Ledger documentation warns that its filters may be incomplete. What holds it off A is that the boundary is not drawn where a buyer needs it. No firm size or segment is stated anywhere on the vendor's own surfaces, no practice area is identified as unsupported, and government use is not addressed. More sharply, the coverage claim itself is inconsistent across live surfaces: the lawyer page and FAQ describe 3.5 million official texts across eight jurisdictions and name them as France, the EU, Germany, Italy, Spain, Belgium, Switzerland and Luxembourg, while the source documentation describes France, the EU, Luxembourg, Great Britain and Switzerland, carries no German, Italian, Spanish or Belgian corpus at all, and its own metadata says two million references. Verified 12 September 2026.