Product and intended use
Define the configuration, people, information and tasks being assessed.
Plan, contracting entity and modules
Public claim only
The contracting entity on the public website Terms is Draftwise Inc., registered in Delaware with a New York office. Those Terms cover the Site, not a published product MSA. Marketed modules include the Word add-in, Knowledge Console, Playbook Studio, Deal Table and Legal Ontology. Trust FAQ (read 2026-09-20): AI features are not available on customer-existing-cloud or on-premises installs.
Ask: Confirm the contracting entity, which modules are on our Order, whether AI is in scope for the deployment we would buy, and whether we are on a negotiated MSA/DPA rather than the website Terms of Use.
Deployment model
Public claim only
Site FAQ and Trust FAQ describe four patterns: (1) containerised multi-tenant SaaS; (2) a private tenant or private cloud in a region of the customer's choosing (site FAQ names AWS, Azure or GCP, with customer-held keys on private cloud); (3) install into the customer's existing cloud (Helm, VM images or Linux binaries), with AI features not available; (4) on-premises, also without AI features. Default SaaS region is not named.
Ask: Which of SaaS, private tenant, customer-cloud or on-premises applies to our Order; in which cloud account and region; who holds keys; and is generative AI actually enabled on that pattern?
Product features in scope
Public claim only
Product pages describe Word-native drafting and redlining from the firm's deal history, automated playbooks and collections, Deal Table (corpus extraction, obligations, market patterns), Knowledge Console (collections, users/licences, usage analytics, playbooks) and a Legal Ontology. Agentic copy claims deep DMS research, interpretation of redlines and comments, and editing entire documents. EDGAR comparison is listed.
Ask: Which of Word drafting/review, Playbook Studio, Deal Table, Knowledge Console, Legal Ontology, EDGAR compare and agentic document-edit will be enabled and billed on our Order?
Governing law and Australian coverage
Public claim only
Website Terms of Use select New York law and exclusive New York courts, with AAA arbitration after 30 days' informal negotiation. That instrument is the Site terms. No public product MSA, Australian law option, or Australian office was found. Named customers on the site and in Legal IT Insider are US, UK and New Zealand firms. EDGAR is a US securities corpus.
Ask: Will the product MSA accept Australian law and courts (or a non-exclusive Australian forum), and what Australian-law coverage, if any, exists beyond EDGAR and the firm's own documents?
Data, controls and requirements
Compare the actual configuration with firm, client and jurisdiction requirements, supported by evidence.
Client-facing use and authority to process
Public claim only
The public product is lawyer-facing (Word add-in and admin console), not a client chatbot. Privacy policy assumes access to the firm's employees, systems and proprietary content and states customers agree to the policy by using the services. No template Australian engagement-letter disclosure, outside-counsel wording, or technical client/matter exclusion control was found. Privacy also states typical deploy is behind the customer's firewall and that Draftwise 'processes no personal data', which sits uneasily with a SaaS Word add-in that reads contracts.
Ask: Confirm there is no client-facing AI on our plan; provide wording covering hosting countries, model providers, retention and training for engagement terms; and can we exclude individual clients or matters technically?
Foundation-model providers
Public claim only
A 12 February 2026 vendor blog states every LLM provider is on an enterprise zero-retention contract and that requests can be routed across interchangeable providers. It does not name those providers. A Microsoft customer story describes Cohere Command, Embed and Rerank plus o-series reasoning models on Azure AI Foundry, including reinforcement fine-tuning on Draftwise proprietary data. Customer ability to pin or disable a provider is not documented.
Ask: Name every model family, version and hosting route that will process our content on the deployment we would buy; whether we can pin or disable a provider; and produce the current ZDR/DPA artefacts for each.
Subprocessors
Not confirmed
No dated public subprocessor list or DPA URL was found. Privacy policy lists categories only: cloud hosting, payment/billing, communications/support/analytics, security/identity, and professional advisors. Site FAQ names AWS CloudWatch for logs and AWS, Azure and GCP as cloud partners. Trust Center resources include a Third-Party Management Policy behind request-access.
Ask: Provide the current dated list with legal entity names, countries of processing, data types each receives, 30-day (or longer) change notice, and a right to object or terminate if a new subprocessor is unacceptable.
Training and secondary use
Not confirmed
Homepage: 'Your data never trains public models.' Site FAQ and the February 2026 ZDR blog: Draftwise will not share or reuse firm data in any form, including trained intelligence or aggregates; LLM contracts forbid training and fine-tuning on customer inputs. Privacy policy: Draftwise may aggregate and/or de-identify information collected through the services and website and 'may use aggregated data for any purpose, including without limitation for research and marketing purposes.' The Microsoft story describes reinforcement fine-tuning on Draftwise proprietary data, which is not the same as customer matter files but is not defined in a customer contract we could read.
Ask: Confirm in the signed Order that prompts, uploads, outputs, embeddings, logs, feedback and telemetry are not used to train or improve any model for other customers, and that aggregated or de-identified data cannot be reverse-engineered to our matters; reconcile the privacy-policy aggregate-use language with the FAQ/ZDR 'no aggregates' claim.
Feedback, ratings and product improvement
Public claim only
Website Terms treat questions, comments, suggestions, ideas and other Site 'Submissions' as non-confidential and as Draftwise's sole property for unrestricted use. Knowledge Console includes usage analytics (adoption, engagement, usage trends). No documentation of in-product thumbs, sampled QA of matter content, or an opt-out for product-improvement telemetry on the Word add-in was found.
Ask: Can in-product ratings, telemetry and QA sampling of prompts/outputs be disabled for our tenancy; what exactly is stored in usage analytics; and is any of it used to train models or shared beyond our tenant?
Retention of prompts, documents and logs
Not confirmed
ZDR blog (12 February 2026): inputs and outputs are not stored beyond the LLM processing window; only operational metadata such as latency may be retained at the provider. Trust FAQ: AWS CloudWatch logs retained up to two years. Data Retention Policy: Draftwise retains data it holds for no longer than five years unless law requires longer. Privacy policy's 'How long do we store your data?' heading does not state a period. Administrator-configurable workspace retention was not documented. Five-year company retention and LLM ZDR can both be true if they apply to different stores; the public documents do not map stores.
Ask: For our tenancy, list retention for prompts, document copies, embeddings, playbooks, ontology, Deal Table extracts, CloudWatch logs, backups and each LLM provider, and whether ZDR is contractual for every model actually used.
Deletion on instruction and on exit
Public claim only
Trust Center control text includes 'Customer data deleted upon leaving' and 'Data retention procedures established'. The Data Retention Policy says data is deleted at the end of the applicable period. Privacy policy offers a right to request erasure of 'certain data' without a numbered SLA. No matter-level self-service deletion, backup lag, subprocessor flow-down or deletion certificate was found in a readable document.
Ask: Contract a deletion and return period (including backups, embeddings, ontology, Deal Table extracts and subprocessors), matter-level deletion, and a written certificate; confirm the period after termination during which we can export first.
Vendor and contractor access to content
Public claim only
Site FAQ: on private cloud, Draftwise retains limited access to facilitate support; on-premises, the support team requires limited access for routine upgrades and 24/7 technical support. Privacy policy lists professional advisors (auditors, accountants, outside counsel) among recipients of personal data. Customer-visible logs of vendor access were not found. LLM human-review settings are not described.
Ask: Enumerate every human-access path (support, QA, abuse, legal process) at Draftwise and each model provider; can we require per-ticket consent; is access logged and exportable; and can human review of content be switched off?
SSO, MFA, admin controls and audit logs
Public claim only
Knowledge Console claims central user, licence and self-serve registration management plus usage analytics. Homepage and ZDR blog claim DMS permission mirroring. Trust FAQ: CloudWatch logs with near-real-time collection, retained up to two years. No public documentation of SAML/OIDC SSO, MFA enforcement, SCIM, customer-exportable prompt-level audit logs or SIEM streaming was found.
Ask: On the plan we are buying, confirm SAML/OIDC SSO-only, SCIM deprovisioning, MFA, per-feature admin switches, and an exportable audit log of prompts, exports, admin changes and vendor access with stated retention.
Encryption in transit and at rest
Public claim only
Site FAQ and Trust FAQ: AES-256 at rest; HTTPS with TLS 1.2 or 1.3 in transit. Private-cloud FAQ: the firm IT maintains encryption keys. SaaS key management (KMS, HSM, customer-managed keys), coverage of embeddings/logs/backups, and rotation are not named in a readable contractual document. Trust Center lists a Cryptography Policy behind request-access.
Ask: Name algorithms and protocols for every store (including embeddings, logs, backups and the LLM link), who holds keys on our deployment, rotation, whether customer-managed keys are available on SaaS, and the SOC 2 control references.
Independent security assurance
Public claim only
Trust Center lists SOC 2 Type II (filename Draftwise - SOC 2 Type II - 2025.pdf), ISO/IEC 27001:2022 official certificate, a penetration-test pack and a 22 January 2026 letter of attestation, GDPR Compliance Policy, cyber-insurance certificate (2026), and ISMS procedures — all request-access except where marked viewable; none of the reports were inspected. Trust FAQ also claims Cyber Essentials Plus; that certificate is not in the public resource list. ZDR blog says ISO 42001 aligned, not certified. IRAP was not found.
Ask: Provide under NDA the current SOC 2 Type II (scope, period, exceptions), ISO 27001 certificate, Cyber Essentials Plus certificate, the 2026 penetration-test summary covering AI features, and any ISO/IEC 42001 or Australian Essential Eight/IRAP mapping.
Contract, DPA, confidentiality, IP and liability
Public claim only
The only public terms are website Terms of Use (New York law, AAA, vendor may assign, Site Submissions become Draftwise property, no product SLA). No standalone DPA, APP schedule, Output IP clause, privilege clause or liability cap for the product was found. Privacy policy certifies to the EU-U.S., UK and Swiss Data Privacy Frameworks. Trust Center GDPR policy is request-access.
Ask: Will you sign an APP-aligned DPA with 72-hour breach notice, flow-down to every subprocessor, customer ownership of inputs/outputs/playbooks/ontology, Output IP indemnity, Australian law, and confidentiality language that can meet NSW SC Gen 23 cl 9A for court-protected material?
Cross-border processing and Australian residency
Public claim only
HQ is New York. Privacy policy is a US DPF certification for personal data received from the EU, UK and Switzerland. Trust FAQ: private tenant or private cloud 'in a region of your choosing'; site FAQ names AWS, Azure or GCP. Default SaaS region is unpublished. No advertised Australian region or IRAP statement. If personal information in matter files is processed in the US (or another overseas region), APP 8 analysis is required.
Ask: Is an Australian region available for storage and inference on the AI-enabled deployment we would buy; if not, list every category of data that leaves Australia, the US CLOUD Act exposure, and the APP-equivalent contractual commitments we can rely on?
Breach notification and incident history
Not confirmed
Trust Center lists an Incident Response Plan behind request-access; it was not read. The public privacy policy does not state a numbered notification period, a suspected-incident trigger, or Notifiable Data Breaches cooperation. No public status page or incident register was found. Security contact is security@draftwise.com.
Ask: Contract notification within a defined period (for example 72 hours) of suspected incidents affecting any Customer Data, cooperation with NDB assessments, a named security contact with Australian-hours coverage, and disclose material incidents in the last 24 months.
Change of terms, models and control
Public claim only
Website Terms allow Draftwise to change, suspend or discontinue the Site without notice and to assign any rights and obligations at any time. Privacy policy may be amended from time to time by posting. The ZDR blog describes vendor-controlled multi-provider routing if a provider's posture changes. No contractual advance notice of model-version swaps, a pin/testing window, or change-of-control termination for the product was found. Last published equity round found is the March 2024 Series A.
Ask: Lock the Order against mid-term term changes; require written notice and a testing window before model, prompt or retrieval changes; and give us termination and data-return rights on change of control, discontinuance or a prohibited-provider substitution.
Export, transition and shutdown
Public claim only
Trust Center claims customer data is deleted upon leaving. ZDR blog: the customer retains full ownership and control of its data. No bulk-export API, open-format dump of playbooks/ontology/Deal Table, post-termination read-only window, transition assistance rate card or escrow was found. Website Terms disclaim liability for loss of Site data.
Ask: Demonstrate a full tenancy export (documents, playbooks, ontology, Deal Table extracts, prompts, outputs, audit logs) in open formats, a 30–90 day retrieval window, deletion including derived artefacts with a certificate, and transition assistance rates.
DMS and storage connectors
Not confirmed
Product and Deal Table pages list iManage, NetDocuments, SharePoint and Google Drive. Product copy: from the moment you connect the DMS, Draftwise auto-curates collections. A 3 September 2026 vendor blog argues the DMS must remain the only permission source and that access should be checked at prompt time against live DMS controls, with blocklists and audit events written back to the DMS. Those two descriptions (ingest-on-connect versus query-time-only) are not reconciled in a technical specification we could read.
Ask: For each connector we would enable, show the identity used, scopes, whether a service account exists, whether documents are copied or only queried, how wall changes propagate, and how we disconnect and purge connector copies.
EDGAR and other public retrieval
Public claim only
Homepage and product logos market comparison of revisions against EDGAR. It is not described as a general open-web browse. Whether the open contract is sent to EDGAR or a third-party EDGAR API, whether the feature can be disabled per tenancy, and citation behaviour when nothing is found were not specified in a contractual document.
Ask: Can EDGAR (and any other web retrieval) be disabled for our tenancy; what query payload leaves the tenant; and does the product refuse to answer when no source is retrieved?
Agent actions and confirmation gates
Not confirmed
Product page: agentic AI performs deep research on the DMS, interprets redlines and comments, and edits entire documents in minutes. Playbook Studio applies automated redlines from generated rules, then the user can audit citations, edit and export. Deal Table continuously merges contract data from connected stores. No public list of actions that require per-action confirmation, write-back controls, or prompt-injection testing was found. Trust FAQ: AI features are off on some deployment patterns.
Ask: List every action the agent can take on Word, iManage, NetDocuments or email; which require explicit per-action approval; how untrusted document text is isolated from instructions; and provide the latest injection/red-team summary.
Human review of outputs
Public claim only
Playbook Studio describes automated redlines followed by lawyer audit of citations, manual edits in the file, and export. Drafting is inside Word, where a lawyer can accept or reject changes. There is no documented mandatory approval gate that ordinary users cannot disable, no junior-user supervisor sign-off, and no exportable per-output review log with model version.
Ask: Can we require supervisor sign-off before redlines leave the firm, export a per-document record of model version and reviewer, and disable any workflow that applies changes without a human click?
