Dodonai

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.

Every purchasable function is a model output. Deposition summarisation, medical chronology construction, OCR of scanned records including handwriting and tables, extract and draft agents, and semantic search that finds evidence by meaning rather than keyword are the whole of the platform, and the pricing page charges by pages processed with multipliers by process type, which is charging directly for inference. Remove the models and there is a document store with a search box. The company describes itself in its own FAQ as an AI-powered document processing platform built for legal professionals. 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.

Grounding is real, documented and unusually concrete for this product class. Deposition summaries are described as page-and-line cited, meaning each summary point carries a pointer back to the transcript location it came from, and the transcript product adds automated Blue Book citations and an assistant that works over the full text. That is a retrieval mechanism a reader can understand and a reviewer can check against the source. Failure modes are acknowledged in the agreement rather than only in marketing: section 3(d) states that given the probabilistic nature of machine learning the Services may produce incorrect Output that does not accurately reflect real people, places or facts, and advises evaluating accuracy including by human review. What is missing is measurement. No accuracy figure is published, no test set is described, and the home page FAQ item asking how Dodonai ensures the accuracy of AI-generated summaries did not render in this pass and is not credited.

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.

Oversight is asserted without a described mechanism in the platform itself. Section 3(d) of the terms advises that the customer should evaluate the accuracy of any Output as appropriate, including by using human review, which is guidance rather than a control. The one firm commitment located sits outside the product: the AI Managed Services arm is described as operating with an attorney in the loop on every output, which is a services promise rather than a platform behaviour. Every action is stated to be logged, which is an audit surface rather than a review gate. Nothing published describes what the system does unattended against what a person approves, no threshold is stated, and no route back to human judgement is set out for the self-serve platform. Searched the home page, the terms and data processing addendum, and the pricing page on 4 September 2026.

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.

An unattributed testimonial stands in for deployment evidence. One quotation is published, from a first name and a job description, Ben, Commercial Litigator, saying that on the first day of use Dodonai summarised all the deposition transcripts from a case in seconds, saving hours of time and hundreds of client dollars. No organisation is named anywhere, there is no logo strip, no case study, and no customer count. The claim of hundreds of client dollars saved is a result quoted with no basis, no matter size and no date. Comparison pages exist for deposition and chronology software but are the vendor's own category content rather than deployment evidence. Searched the home page, the pricing page and the terms on 4 September 2026.

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 privilege limb is met more explicitly here than almost anywhere in this pull, and other limbs are what hold the grade. Section 3(e) states that both parties acknowledge Content may include confidential and privileged information, that Dodonai will use all reasonable efforts to keep it confidential, and then goes further than any peer: all parties agree that both the Input and the Output constitutes attorney work product and that Dodonai shares a common interest with its users. Segregation is documented as isolated per-client workspaces with case data siloed, role-based access so team members see only what they need, and every action logged. The position on model providers is stated in section 3(c): Dodonai only works with LLM Partners that disclaim use of the Content for developing or improving their own services and covenant to delete customer data after a reasonable time, given as 30 days for OpenAI. What holds this at B is the rest. Training by Dodonai itself is permitted on opt-in rather than prohibited. No retention period for Dodonai's own storage is published anywhere, only deletion on termination or reasonable request. And only one LLM partner is named, with the phrase including OpenAI leaving the others unidentified.

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.

Nothing published addresses the advice line, and the grade rests on documents that were read rather than on an untested gap. The full terms and the data processing addendum were retrieved on 4 September 2026 and contain no statement that the Services are not legal advice, no competence or supervision language, no reference to any rule of professional conduct, and no jurisdiction limit. The nearest provisions point the other way: section 3(e) characterises Input and Output as attorney work product and asserts a common interest with users, which presupposes a legal context without addressing the vendor's position in it. The audience includes people who are not lawyers, with published material for paralegals, court reporters, independent medical examination companies and IME doctors, and section 1 sets the minimum age at 13 with parental permission under 18, which is unusual language for a product processing privileged case files. Section 3(d) advises human review of Output but frames it as an accuracy precaution rather than a professional responsibility one.

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 position on the limits of the AI is published without any mechanism behind it. Section 3(d) of the terms acknowledges the probabilistic nature of machine learning and the possibility of incorrect Output, and section 3(b) explains that output may not be unique across users. That is a principle rather than a governance regime: nobody inside Dodonai is named as accountable for model behaviour, no pre-release evaluation is described, no testing results are published, and there is nothing at all on bias or uneven output, which matters on a product that builds medical chronologies and summarises testimony. A Safety and Trust page exists under the AI Managed Services section and was not opened in this pass; it is named here as an unread surface rather than credited, since crediting a page by its title is not evidence of its contents.

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.

Substantive published policy covering most of the ground. Access control is specific: AES-256 at rest, TLS 1.2 or higher in transit with documents described as encrypted from the moment they leave the browser, isolated per-client workspaces, role-based access, and every action logged. Section 5(a) of the data processing addendum enumerates the security measures required across personnel, facilities, hardware and software, storage and networks, access controls, monitoring and logging, vulnerability and breach detection, incident response and encryption. Incident practice is committed at section 5(c), requiring notification of any personal data breach by Dodonai, its LLM Partners or other third parties acting on its behalf without undue delay, with assistance in investigation under section 3(a). Deletion is addressed at section 8, on termination or upon reasonable request. Two elements are missing: no retention period is published for Dodonai's own storage during the term, and no subprocessor list is published, with LLM Partners identified only as including OpenAI.

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. Section 7(b) puts the Services on an as-is basis and disclaims all warranties express, implied or statutory, expressly including that the Services will be accurate or error free or that Content will be secure or not lost or altered, so no warranty is given that a buyer could invoke. Section 7(a) runs in one direction only: the customer indemnifies Dodonai, and no indemnity is given to the customer for anything, including third-party intellectual property claims. Section 7(c) caps aggregate liability at the greater of twelve months of fees or one hundred dollars, and excludes all indirect and consequential damages including loss of data. No insurance position was located. What keeps this off the floor is that the allocation is published and readable before signing rather than absent, and that the cap carries a stated floor; on the question this axis asks, the published answer is that the customer bears the loss entirely.

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 practice systems was located. No document management system, case management platform, email client, word processor or e-signature product is named anywhere on the surfaces read, and no integrations page or developer documentation exists in the navigation or footer. The only connectivity evidence is indirect: section 2(c)(iv) of the terms refers to extraction of data through the API as permitted, which establishes that an API exists without describing what it connects to, and the enterprise pricing tier lists custom integrations as an unspecified inclusion. Two surfaces that might bear on this were not read and are named so the limit is visible: the home page FAQ item asking how Dodonai integrates with an existing workflow did not render, and the Law Firm Case Management and Document Management pages under Industries were not opened. 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.

The tenancy limb is published and the residency limb is entirely absent, which is the documented gap on this axis. Isolated per-client workspaces are stated, with case data siloed, role-based access limiting what team members see, and every action logged, which tells a buyer how customers are separated from one another. Nothing states where data is stored or processed: no region, no country, no hosting provider and no processing location appears on any surface read, and the terms note only that the Services may be used in geographies currently supported by Dodonai without saying which those are. Publishing tenancy alone clears the band below, where neither limb is stated, and this record sits in the space the operator has logged as fitting neither adjacent band cleanly.

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.

Certification is stated and repeated as a product claim rather than shown as a decorative badge, appearing in the hero, in a dedicated security section and in the platform description: SOC-2 certified and HIPAA fully compliant, with a Business Associate Agreement available on request. A route to evidence exists contractually, since section 1(i) of the data processing addendum commits Dodonai to cooperate with customer assessments and audits where required by law, or alternatively to make available a summary of third-party audit results or certification reports. It stops short of the top on every accessible-evidence limb. The claim is written as SOC-2 without specifying Type I or Type II, which is a material difference and is not resolved anywhere; no certifying body or auditor is named; no scope, observation period or report date is published; there is no trust portal; and penetration testing is not mentioned. On the third-party verifiability test a buyer cannot check the claim against an auditor's register without contacting Dodonai.

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 supply chain is partly disclosed, with one provider named and change notification committed in unusual detail. Section 2(c) requires customers to adhere to the terms and policies of any LLM Partners used for data processing, including OpenAI, and links OpenAI's policy pages; section 3(c) states that Dodonai only works with LLM Partners that disclaim use of Content to develop or improve their own services and that covenant to delete customer data after a reasonable time, given as 30 days in OpenAI's case. Change notification is stronger than most records carry: section 1(g) of the data processing addendum commits to notifying customers of additional LLM Partners, gives a fifteen-day window to object on reasonable grounds relating to data protection, sets out four cure options including declining to use the partner or offering an alternative, and gives a termination right with a refund of prepaid fees if the objection is not resolved within thirty days. What is absent is the rest of the picture: no model is named, the phrase including OpenAI leaves the other partners unidentified, no subprocessor list is published, and where inference runs is not stated.

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 learn the whole cost structure without speaking to anyone. Four tiers are published with rates: 25 dollars a month billed annually at 300 for 4,800 credits a year, 50 a month billed annually at 600 for 24,000 credits, 83 a month billed annually at 1,000 for 120,000 credits, and a custom enterprise tier. The unit is defined rather than gestured at, with one page stated as approximately 400 tokens of text, and a table gives the effective cost per page at each volume by process type with the multiplier shown: standard summaries at 1x running from 6.3 cents to 0.8 cents, custom summaries at 2x, advanced analysis at 3x, and OCR and transcription free at every tier. Every feature is stated to be included in every plan with no per-seat fees and no feature gates, and the nine included capabilities are itemised. What the enterprise tier adds is named as custom integrations, SLA guarantees and dedicated onboarding. A seven-day, hundred-page free trial is published with self-serve signup, and section 4(c) of the terms commits to fourteen days notice before a price increase takes effect.

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.

Coverage is described with real substance across three dimensions and the boundary is left open. Six practice areas each carry their own page: personal injury, medical malpractice, mass tort, product liability, workers' compensation and disability benefits. Eight use cases each carry their own page, including medical record review, IME reports, demand letters, expert witness preparation and contract review extraction. Six industries are addressed separately, extending beyond law firms to court reporters, IME companies, IME doctors and medical record retrieval businesses, which is an unusually precise statement of who else buys this. What is missing is the edge: no firm size or segment is stated, in-house and government use are not addressed, and nothing says which matter types or document types the product does not handle. The practice area set is plaintiff-side almost throughout, which is a coverage fact rather than a limit the vendor states.