Draftwise
Contract drafting, review and negotiation AI that works from a firm's own precedent rather than from a generic corpus, operating inside Microsoft Word. Draftwise connects to the firm's document management system, reads its complete deal history including documents, clauses and tags, and uses it to draft new agreements, interpret incoming redlines and comments, follow version history, and edit whole documents. Playbooks are generated automatically from existing precedent and guidance instead of being written by hand, and contract collections are curated in the background once the DMS is connected. On the negotiation side it lays out comparable terms from past deals by client, counterparty or industry, and checks proposed positions against EDGAR filings so a position can be shown to be market-tested. The Legal Ontology platform, announced in August 2026, adds a structured layer over the same material that makes positions, fallbacks and counterparty patterns queryable across every deal a firm has done. Named integrations are Microsoft Word, iManage, NetDocuments and EDGAR. Deployment is unusually flexible for this market: a private cloud in the customer's choice of AWS, Azure or GCP where the firm's own IT holds the encryption keys and audit control and no data leaves the VPC by default, deployment into a firm's existing cloud through a helm chart, VM images or Linux binaries, or an on-premise install. Draftwise Inc. is an independent company incorporated in Delaware with its registered office in New York, and is certified under the EU-U.S. Data Privacy Framework together with the UK Extension and the Swiss-U.S. framework. Customers named on the site include Orrick, Gunderson Dettmer, Mishcon de Reya, Katten, McGuireWoods, Womble and Maddocks.
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.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
The models are the product and there is no conventional system underneath, because the system of record belongs to the customer. Draftwise connects to the firm's existing document management system and the machine learning is what turns that corpus into work: the product page describes agentic AI that performs deep research across the DMS, interprets redlines and comments, understands version history and edits entire documents, analyses millions of documents in real time to select relevant precedent, builds playbooks automatically from existing precedent and guidance rather than by hand, and auto-curates contract collections in the background with no setup. The Legal Ontology platform announced August 2026 is a structured intelligence layer over the same material. Remove the models and what remains is the firm's own DMS, which Draftwise does not sell.
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 repeatedly and measured nowhere. The claims are confident and unquantified: lawyer-quality edits described as precise and accurate, drafting delivered with total confidence, and the Legal Ontology page headed with answers you can trust. Searched the home page, the product page, the FAQ, the terms of service and the privacy policy on 31 Aug 2026 and located no accuracy figure, no hallucination rate, no benchmark, no described test set and no published evaluation of any kind. What does support the output is provenance rather than measurement: drafting is grounded in the firm's own precedent, which a lawyer can open in their own DMS, and negotiation positions are checked against EDGAR filings, a named public corpus. No retrieval method is described, no citation mechanism is documented, and nothing addresses what the system does when the precedent does not support a position.
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.
Autonomy is claimed prominently and no oversight mechanism is published. The product is described as a personal drafting agent that drafts, reads, redlines, researches and reasons like the buyer's best team member, as agentic AI built to draft, and as editing entire documents in minutes, with collections curated in the background requiring no setup and no maintenance. The one real review surface is implicit rather than described: the product operates as a Microsoft Word add-in, so redlines land in a document a lawyer opens, and the vendor says they are easy to review and negotiation-ready. Searched the home page, the product page, the FAQ, the terms and the privacy policy on 31 Aug 2026 and located no statement of what the system does unattended, no threshold at which it stops, no described approval or review point, and nothing on what happens after an output is wrong.
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 unusually deep roster of named customers speaking on the record, with no figures attached. Attributed quotes come from Vedika Mehera, Director of Orrick Labs, and Wendy Butler Curtis, Chief Innovation Officer at Orrick; Naveen Pai, Chief Knowledge Officer, and Joe Green, Chief Innovation Officer, both at Gunderson Dettmer; Oliver Neasham, Managing Associate at Mishcon de Reya; Cameron Reeves, Partner at Mayne Wetherell; and Franck Sekri, Partner at Sekri Valentin Zerrouk. Zac Padgett, a partner at Orrick, describes a specific task in detail, reviewing comparable deal documents and surfacing a market-standard percentage during the closing hours of a deal. A 14-logo strip adds Katten, McGuireWoods, Womble, Chapman, Maddocks, Borenius and Bronstein Zilberberg, and dedicated case study pages exist for Orrick, Gunderson Dettmer and Mishcon de Reya. What is missing is measurement: every quote is qualitative, no figure of any kind is attached to any customer on the pages read, and no deployment date or method appears. The case study pages were not opened on 31 Aug 2026, so the figures and dates limb is rebuttable.
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.
Strong architecture, strong assertions, and nothing a buyer can hold the vendor to. The assertions are specific: the FAQ states that the firm retains complete control of its data and that Draftwise will not share or disclose any data or firm intellectual property in any form, including trained intelligence or aggregates, which is an unusually explicit formulation. The architecture backs it, since private cloud and on-premise deployments leave encryption keys and audit control with the firm's own IT and no data leaves the VPC by default. Three things hold this at C. The only agreement published on the property is a website terms of use, last updated 11 August 2025 and scoped to the Site, so there is no customer agreement a buyer can read before signing; its clause 8 instead permits Draftwise to access, store, process and use information provided, and clause 9 makes Submissions non-confidential and Draftwise's sole property. The no-training statement is qualified on its face to public models, which leaves the vendor's own models unaddressed on a product built to learn a firm's positions. And privilege and work product are not addressed anywhere on any surface read on 31 Aug 2026, on a product whose corpus is a law firm's client matter files.
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 on the advice line. Searched the home page, the product page, the FAQ, the terms of service, the privacy policy and the trust centre landing page on 31 Aug 2026. No statement that Draftwise is not a law firm and does not provide legal advice was located, no professional responsibility or ethics page exists, no bar or ethics guidance is named including ABA Formal Opinion 512, and nothing addresses a lawyer's competence or supervision duties when using the tool. The website terms of use disclaim warranties and accuracy in general commercial language but say nothing about legal advice. The buyer here is a lawyer rather than a consumer, which lowers the unauthorized practice exposure relative to a public-facing product, but the axis asks what the vendor has published about the line between tooling and advice, and the answer is nothing.
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 for model behaviour was located. Searched the home page, the product page, the FAQ, the terms of service, the privacy policy and the trust centre landing page on 31 Aug 2026. Nothing names an internal owner accountable for model behaviour, describes what is tested before release, publishes a responsible AI framework, or discloses anything about uneven output across matter types, counterparties or practice areas. The home page carries a tab labelled Ethics, which is the closest thing on the property, and its content is a single confidentiality claim that customer data never trains public models. Confidentiality is a different subject from governance over model behaviour and does not carry here. The published Information Security Policy and the security certifications likewise address information security rather than AI governance.
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 material with two specific gaps. Draftwise publishes an Information Security Policy and a Data Retention Policy as separate documents alongside the privacy policy, which is more than most of this market offers openly. The FAQ gives real operational detail rather than assurances: AES-256 at rest and HTTPS over TLS 1.2 or 1.3 in transit, audit logging through AWS CloudWatch collected in near real time with rules-based and composite alerting and anomaly detection on containers and applications, and logs retained for up to two years. Draftwise Inc. is certified under the EU-U.S. Data Privacy Framework with the UK Extension and the Swiss-U.S. framework, with JAMS named as the independent recourse mechanism and FTC enforcement jurisdiction acknowledged. Against that, no named subprocessor list exists anywhere; the privacy policy lists categories of recipient only. No incident response or breach notification practice was located on any surface. And the privacy policy's own section headed how long data is stored contains the text of the following section on changes to the policy instead of a retention answer, so the document that should answer retention does not. The separate Data Retention Policy was not opened on 31 Aug 2026, so that last point is rebuttable.
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 limitation clause, and the clause governs the website rather than the product. The document published as Terms of Service, last updated 11 August 2025, is a website terms of use: it is expressly made concerning access to and use of the Site. Within it, clause 17 disclaims all warranties including any representation about accuracy or completeness, clause 18 excludes liability for direct, indirect, consequential, exemplary, incidental, special and punitive damages arising from use of the Site with no cap stated, clause 19 requires the user to indemnify Draftwise, and clause 20 disclaims all liability for loss or corruption of transmitted data. No indemnity running to the customer, no warranty on output and no insurance position was located. No master subscription agreement or product terms were located anywhere on the property, so the enterprise buyer's actual allocation of loss is negotiated rather than published. This is a C rather than a D because a limitation of liability is published and readable in advance, and clause 17 extends its disclaimer to the services as well as the Site.
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.
Real integrations into the systems that matter for this buyer, named but not documented. The product page names iManage and NetDocuments alongside Microsoft Word, EDGAR and a Microsoft Partner mark, which is the right set for a product sold to transactional law firms, and the function is described rather than merely asserted: Draftwise connects to the firm's complete deal history including documents, clauses and tags, performs deep research on the DMS, follows version history, and auto-curates contract collections from the moment the DMS is connected. The privacy policy confirms the product operates as an installable Microsoft Word add-in. What is absent is depth an implementer could use: no integrations page, no statement of what syncs in which direction, no description of what a firm must configure, and no developer or API reference index was located on 31 Aug 2026. One detail on that page is worth a buyer's notice: one logo in the integrations strip is an unreplaced placeholder, published with the filename and alt text reading simply Something.
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 most complete deployment disclosure read in this pull, published openly on the FAQ rather than held for a security questionnaire. Three routes are described with the trade-offs spelled out. A private cloud dedicated to the firm in the customer's choice of AWS, Azure or GCP, where the firm's own IT maintains encryption keys and audit control, Draftwise retains only limited access to support the deployment, no data leaves the VPC by default, and network access can be restricted to the firm's IP subnets. Deployment into the firm's existing cloud infrastructure through a helm chart and private image registry, VM images or Linux binaries, administered either by Draftwise or self-serviced by firm IT. Or an on-premise install, again with limited vendor access for upgrades and support. The privacy policy adds that products are typically deployed behind the customer's firewall on servers the customer's IT department manages, and that Draftwise consequently processes minimal information in its own systems. Residency follows from the customer's choice of provider and estate rather than from a vendor region list. One tension a buyer should resolve: if no data leaves the VPC by default, where the language models run relative to that boundary is not explained anywhere, and no model provider is named.
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.
Certifications are claimed, a trust centre exists, and the property contradicts itself on what is actually held. The home page states that Draftwise is SOC 2 Type II and ISO 27001 certified and GDPR compliant. The FAQ, on the same property, states that Draftwise has been certified for Cyber Security Essentials Plus, a UK cybersecurity standard, and is SOC 2 compliant. Those are three separate discrepancies: ISO 27001 appears on one page and not the other, Cyber Essentials Plus appears on the other and not the first, and compliant is not the same claim as certified. A Vanta-hosted trust centre at security.draftwise.com is linked openly from the footer and is credited here as reachable, since its body renders client side and could not be read by the index, which is a retrieval limit on this side rather than a gap on the vendor's. What could not be established from any readable surface on 31 Aug 2026 is the substance: no auditor is named, no coverage period or report date is given, no audit scope is described, and although the FAQ asserts routine security evaluation it names no penetration testing partner and publishes no summary. Further encryption and audit questions are directed to a sales address.
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.
Nothing published about the model supply chain a customer inherits. Searched the home page, the product page, the Legal Ontology summary, the FAQ, the terms of service and the privacy policy on 31 Aug 2026. No model provider is named, no model or version is identified, no architecture is described beyond the phrase agentic AI built to draft, no subprocessor list exists, and there is no commitment to notify customers when any of it changes. The only reference to underlying models anywhere is the confidentiality claim that customer data never trains public models, which acknowledges that third-party models exist without saying whose are used, where they run, or what they see. That gap is sharper here than it would be elsewhere, because the deployment story turns on data not leaving the firm's own VPC, and a buyer cannot reconcile that with an unnamed external model provider. The FAQ names AWS, Azure and GCP as cloud partners, which is infrastructure rather than model supply.
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. Searched the main navigation, the footer, the home page, the product page, the FAQ, the terms of service and the privacy policy on 31 Aug 2026. There is no pricing page, no tier structure, no per-seat or per-matter unit, no indication of what implementation adds, and every call to action across the property is to book a demo. The website terms of use confirm only that fees may apply and that payment is accepted by credit, wire or ACH in US dollars, with prices subject to change at the vendor's discretion, which is payment mechanics rather than a published rate. Nothing was located that would let a prospective buyer form any view of cost before entering a sales process.
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.
Segment coverage is documented with real precision and the boundaries are never stated. The property segments by firm size with dedicated Big Law and Mid Law pages, and by in-house vertical with six named pages: advertising technology, defense technology framed around FAR and DFARS alongside commercial work, investment funds, life sciences and pharmaceuticals, real estate, and SaaS. The customer roster evidences that breadth rather than merely claiming it, spanning US firms including Orrick, Gunderson Dettmer, Katten, McGuireWoods, Womble and Chapman, and non-US firms including Mishcon de Reya in the UK, Maddocks in Australia, Borenius in Finland, Mayne Wetherell in New Zealand, Sekri Valentin Zerrouk in France and Bronstein Zilberberg in Brazil. Government use is touched only indirectly, through a US Government Rights clause in the website terms addressing FAR and DFARS acquisition and through the defence vertical. What is absent is any statement of where the product stops: no practice areas are excluded, no contract types are named as unsupported, and nothing addresses litigation or any non-transactional work despite the product being explicitly built for transactional practice.
Legal Signals
What each signal meansA 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.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
A public policy or trust page states no training on customer content, with no matching term located in the published agreement.
The home page states that customer data never trains public models and adds that this is not in the vendor's code, and the FAQ states that the firm retains complete control of its data and that Draftwise will not share or disclose any data or firm intellectual property in any form, including trained intelligence or aggregates. Read as published, the commitment is to non-disclosure and to keeping data out of public models; neither formulation addresses training the vendor's own or a firm-specific model, on a product whose stated design learns a firm's positions from its precedent. No training term was located in any published agreement, because the only agreement published on the property is a website terms of use scoped to the Site, whose clause 8 instead permits Draftwise to access, store, process and use information provided.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
Retention is acknowledged in public materials with no stated period.
Retention is acknowledged without a period for customer documents or prompts. The only retention figure located anywhere is for audit and monitoring logs, which the FAQ states are kept for up to two years through AWS CloudWatch. The privacy policy carries a heading asking how long data is stored, but the text beneath it is the text of the following section on changes to the policy, so the document does not answer its own question. A separate Data Retention Policy is published and linked from the FAQ; it was not opened on 31 Aug 2026, so this value is rebuttable on that document. In private cloud and on-premise deployments the customer's own IT holds the data and the keys, which may make retention a customer decision in practice, but no retention control or configurable window is described in vendor material.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
Segregation is asserted in public materials with no published detail on how it is enforced.
Segregation is asserted through the source system rather than through a separate permission model: the home page security card states that Draftwise mirrors the customer's document management system permissions. The claim is more substantiated here than the wording alone suggests, because iManage and NetDocuments are both named as integrations on the product page and the product is described as performing deep research on the DMS and auto-curating collections from it. What is not published is how the mirroring is enforced: no documentation was located on 31 Aug 2026 describing whether the source access model is applied per user at query time or synchronised on a schedule, how matter-level walls propagate, or what happens when firm permissions change. The buyer here includes large transactional law firms, so matter-level walls rather than tenant separation are the relevant test.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Published terms or policy address disclosure to authorities or in response to legal process, and no commitment or reservation regarding customer notice is located anywhere. The vendor has told the customer that data can leave and has said nothing about whether the customer hears of it.
The privacy policy addresses compelled disclosure directly under a heading on disclosure to public authorities, stating that Draftwise may be required to disclose personal data in response to lawful requests including to meet national security or law enforcement requirements, and committing, where permitted by law, to limit such disclosure to what is legally required. That is a minimisation commitment and it is the only obligation the vendor takes on itself. Nothing in the privacy policy, the website terms of use or the FAQ commits Draftwise to notifying the customer that a request has been received, and no transparency report was located on 31 Aug 2026, so whether the firm hears about it remains at the vendor's discretion.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Sources are identified without stating the licence or rights basis.
The corpus is identified because it largely belongs to the customer. Drafting and research run over the firm's own complete deal history, connected through its document management system, with iManage and NetDocuments named as the systems Draftwise reads. The one external source named is EDGAR, the US Securities and Exchange Commission filing system, used to benchmark negotiated positions against market. No licence or rights basis is stated for the EDGAR material and no update cadence or lag is published for either source. The product does not retrieve primary law, so questions of case law and statutory provenance do not arise in the usual form here.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
No located public material addresses whether authority is checked for subsequent history.
Searched the home page, the product page, the FAQ, the terms of service and the privacy policy on 31 Aug 2026. Nothing addresses whether authority is checked for subsequent history, and no citator, treatment signal or currency check was located. The product drafts and negotiates contracts from a firm's own precedent and does not retrieve case law or legislation, so a citator is not part of what it sells. The nearest analogue is the EDGAR benchmarking feature, which tests whether a negotiating position is market-standard against filed agreements rather than whether any authority remains good law.
Refusal and Uncertainty Behaviour
What does the product do when the answer is not in the corpus?
No located public material addresses what the product does when it cannot ground an answer.
Searched the home page, the product page, the Legal Ontology summary, the FAQ, the terms of service and the privacy policy on 31 Aug 2026. No explicit no-answer or abstention path is documented, no confidence or grounding score was located, and nothing states what the product does when the firm's precedent does not support a requested position. Published material runs the other way, describing drafting delivered with total confidence and answers the buyer can trust, without describing the behaviour behind either phrase.
Fabricated Citation Record
Does a public court record exist involving output from this product?
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.
No court order, opinion or disciplinary record naming this product has been located as of 31 Aug 2026. Instrument searched: the AI Hallucination Cases database maintained by Damien Charlotin, which tracks decisions worldwide where a court addressed hallucinated AI content and records the tool implicated where known, searched on both spellings of the product name and on the company name Draftwise Inc., alongside 2026 sanctions trackers and trade press summaries. This is a statement about the public record on the date shown rather than a clearance, and it is bounded by what that database covers. The product drafts and negotiates transactional documents rather than producing court filings, so its output does not ordinarily take the form of citations in a brief.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
No located public material engages with bar or ethics guidance.
Searched the home page, the product page, the FAQ, the terms of service, the privacy policy and the trust centre landing page on 31 Aug 2026. No engagement with any bar or ethics guidance was located, including ABA Formal Opinion 512, US state bar guidance, and Solicitors Regulation Authority or Law Society material despite UK, Australian, New Zealand, French, Finnish and Brazilian firms appearing on the customer roster. No professional responsibility page exists and no ethics opinion is named anywhere. The tab labelled Ethics on the home page carries a confidentiality claim about training rather than any reference to professional obligations.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Public materials claim time savings without addressing billing or disclosure.
Public materials are framed around time saved and lawyer productivity. A named partner describes reducing time spent on preparation, another describes surfacing a market-standard figure within minutes during the closing hours of a deal, the product page describes editing entire documents in minutes, and a case study is titled around improving attorney productivity. Searched the home page, the product page, the FAQ, the terms of service and the privacy policy on 31 Aug 2026 and located no per matter record of AI-assisted work intended for fee purposes and no published guidance on billing, fee or client disclosure treatment. The buyer here is frequently a law firm billing a client, which is the setting where this question bites hardest, and nothing addresses it.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
The material exists behind a sales conversation or an executed agreement.
Two policy documents are published openly and linked from the FAQ, an Information Security Policy and a Data Retention Policy, which is more than most of this market puts on the open web. What a firm would need to satisfy a client AI clause is not there. No subprocessor list was located anywhere; the privacy policy names categories of recipient such as cloud hosting, payment, analytics and professional advisers rather than entities. No model provider is identified at all, so a firm cannot tell its client which providers see matter content. No client-facing consent or notification material was located. A Vanta trust centre is linked from the footer and is the plausible home for the missing artifacts, but its contents render client side and could not be read on 31 Aug 2026, and the FAQ routes further encryption and audit questions to a sales address.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
No located public material addresses court disclosure or verification certification.
Searched the home page, the product page, the FAQ, the terms of service and the privacy policy on 31 Aug 2026 and located nothing addressing court disclosure, AI-use certification or the production of a verification record. Audit logging exists but is infrastructure monitoring rather than a work record: the FAQ describes AWS CloudWatch collecting resource, application and service logs with alerting and anomaly detection, retained up to two years, which is operational telemetry and not a per document account of which model produced which passage and who checked it. No model is identified anywhere on the property, so the model element of any disclosure could not be produced from vendor material in any event.