Contract review and drafting
Spellbook: what to evaluate
Will the redlines follow your playbook and reduce the work needed to finish a contract?
For a commercial-contract team, the useful question is whether suggested changes survive qualified review in the document editor the team already uses. Treat draft quality, missed issues and the effort to produce an approved document as separate measures.
What the vendor says
Spellbook says it redlines contracts against the user's standards and flags terms for attention. Spellbook: review features
The vendor offers drafting from scratch or preferred precedents, and advertises review and drafting within Word and Google Docs. Spellbook: drafting and document editors
Spellbook says its LLM providers do not train on or retain customer data. This homepage assertion does not establish the contractual terms or every processing path. Spellbook: privacy statement
These are attributed product descriptions, not verified performance results.
What remains unresolved
Jurisdiction and availability
Unresolved. England-and-Wales legal accuracy and territorial feature availability have not been verified.
Hosting
Unknown. Storage, inference and backup regions have not been verified.
Training and retention
Unresolved beyond the vendor statement above. Spellbook's own retention, full subprocessors and applicable contract have not been audited.
Editor compatibility
Vendor-described, not functionally tested. Supported versions, plan access and permissions remain unverified.
Your evaluation checklist
Use the following as a proposed test, not a report of tests Margo Legal has run.
Choose authorised, de-identified contracts and have a qualified reviewer record the expected issues before seeing the tool's suggestions. Include both ordinary clauses and an exception the playbook does not cover.
Compare omissions, incorrect findings and unnecessary changes. Check the represented side, defined terms and cross-references, not just whether a redline sounds plausible.
Accept and reject changes in your required editor, then export the document. Confirm that the reviewer can trace each suggestion and that the finished file preserves the intended text.
Measure total review and correction time through to approval. Agree a stop condition for serious omissions before the trial, rather than deciding what counts as success afterwards.
Obtain written terms covering storage and inference locations, subprocessors, training restrictions, deletion and incident notification before using confidential material.
Sources and commercial disclosure
Last source review: . The linked vendor descriptions are not independently tested results. Product capabilities and terms can change.
Margo Legal is developing Margo, a legal-AI offering with potential commercial overlap. These profiles are not neutral third-party endorsements or independently tested comparisons.
Use these pages to prepare questions, not as a recommendation or legal advice.