Reviews
Spellbook Review: AI Contract Drafting Inside Microsoft Word
An independent look at Spellbook, the Microsoft Word add-in transactional lawyers use to draft, redline, and benchmark contracts with GPT-5 and Claude.
Spellbook is a Microsoft Word add-in, not a standalone platform. It uses GPT-5, Anthropic’s Claude, and other large language models to redline, draft, and benchmark commercial contracts directly inside the document a lawyer is already editing, using Word’s native Track Changes. It fits solo practitioners, small-to-midsize transactional firms, and in-house teams handling a steady volume of vendor agreements, NDAs, and commercial contracts. It is not a contract repository, e-signature tool, or post-signature obligation tracker, and it does not do case-law research. Pricing is quote-based rather than published in dollar figures.
Editorial review — Spellbook did not review or approve this article before publication, and this review is based on the company’s own published product and pricing pages, independent trade coverage, and public funding announcements.
Ask a transactional lawyer what slows down a deal and the answer is rarely “research.” It is redlining an NDA for the fourth time this week, checking whether a limitation-of-liability clause matches the firm’s standard position, and drafting a first pass at a services agreement from a client’s half-finished term sheet — all inside Microsoft Word. Spellbook, built by the Newfoundland-founded legal AI company of the same name, is designed around exactly that workflow rather than around a separate research portal or chat window.
What exactly is Spellbook, and how does it work inside Microsoft Word?
Spellbook installs as an add-in inside Microsoft Word for Windows, Mac, and the web, appearing as a sidebar panel alongside the document a lawyer is drafting or reviewing. Rather than asking a lawyer to copy contract text into a separate chatbot, Spellbook reads the open document, applies its suggestions as tracked changes, and lets the lawyer accept, reject, or edit each suggestion the same way they would review a colleague’s redline. The company began in 2019 as Rally, a legal-automation startup founded by CEO Scott Stevenson, before rebranding to Spellbook in mid-2023 after closing a $10.9 million seed round, a shift that Voicebot.ai reported coincided with the company moving from generic legal-task automation to generative-AI contract drafting built on OpenAI’s GPT models. That Word-first design decision, kept ever since, is the main thing that separates Spellbook from browser-based legal AI research tools: the add-in has no separate destination to check — it lives where contract work already happens.
What features does Spellbook include for contract review and drafting?
Spellbook groups its Word add-in around a small set of named tools rather than one general-purpose chat box, which is part of what reviewers point to when explaining why lawyers adopt it quickly. Lawyerist’s independent review, which rates the product 4.1 out of 5, describes it as built specifically for “transactional lawyers who draft and review contracts in Microsoft Word,” and calls out its redlining, playbook, and benchmarking tools as its strongest features.
| Tool | What it does inside Word |
|---|---|
| Review | Redlines a contract using Track Changes, flags aggressive or missing clauses, and suggests alternative language inline. |
| Draft | Generates new clauses or full agreements from a prompt, pulling from a firm’s or team’s own saved precedent language as a starting point. |
| Ask | Answers questions about the open contract — for example, whether an indemnification clause is standard — with citations back to specific provisions. |
| Compare / Benchmark | Checks contract terms against a database the company describes as covering more than 2,300 contract types and market-standard language. |
| Playbooks | Encodes a firm’s or client’s pre-approved fallback positions so the same deviation check runs consistently across every NDA, MSA, or vendor contract. |
| Associate | An AI agent that works across a group of related documents at once — a data room, a set of financing documents, or an employment packet — rather than one file at a time. |
None of these tools attempt full contract lifecycle management. There is no built-in contract repository, e-signature step, or post-signing obligation tracker, which reviewers consistently flag as the trade-off for staying this tightly focused on the drafting and review stage.
What AI models power Spellbook, and what does it actually cost?
Spellbook was originally built on OpenAI’s GPT-4. The company’s current site states the product is now “powered by state-of-the-art LLMs like GPT-5 and Opus,” meaning it routes work across both OpenAI and Anthropic Claude models rather than relying on a single provider. On data handling, Spellbook’s terms of service describe a Zero Data Retention arrangement with those third-party model providers: contract data sent to an underlying LLM is used only in real time to generate the requested output, under agreements that prohibit the model providers from retaining or training on it. The company also states it is SOC 2 Type II audited and supports GDPR, CCPA, and PIPEDA compliance obligations.
On cost, Spellbook does not publish per-seat dollar pricing. Its own pricing page splits offerings into two tracks — one for law firms, one for in-house legal teams — both built around the same underlying “Spellbook Suite” (the Word add-in plus Associate), and both stating that cost “is determined by the number of team members on a license.” Getting an actual quote requires booking a demo or starting a seven-day free trial; there is no self-serve monthly checkout. Prospective buyers should treat any specific per-seat number they see on third-party comparison sites as an estimate, not a published rate, until it is confirmed on a sales call.
How do law firms and in-house teams actually use Spellbook day to day?
Spellbook reports being used by more than 4,500 legal teams across 80-plus countries, with over 10 million contracts processed on the platform and a 4.7-star average rating on G2, according to figures published on its own site. The company’s customer list, drawn from its own case studies and press coverage, spans both corporate legal departments and law firms — it names the international firm Kennedys among its users, alongside in-house teams at companies including Dropbox and eBay. In practice, that usage clusters around a few recurring workflows: an in-house counsel team running every incoming vendor MSA through a Playbook before it reaches a human reviewer; a small transactional firm using Draft to produce a first pass at an NDA or services agreement from a client intake form; and a law firm using Spellbook Associate to work across a bundle of related documents, such as a set of disclosure schedules or employment agreements in a single transaction, instead of opening and reviewing each file separately. The common thread is volume of repetitive commercial paper, not litigation or novel deal structuring, which tracks with how the product itself is scoped.
The company has also raised meaningfully more capital as that usage has grown. After the 2023 rebrand and seed round, Spellbook closed a $20 million Series A in January 2024 led by Inovia Capital, reported at the time by BetaKit, and a further $50 million round in October 2025 led by Khosla Ventures, covered by Artificial Lawyer in an interview with CEO Scott Stevenson. That funding trajectory is a reasonable proxy for staying power: a Word add-in with no repository or e-signature layer is a real bet that the drafting-and-review stage alone is valuable enough to build a company around, and investors have continued to back that bet through multiple rounds.
How does Spellbook compare to Harvey, CoCounsel, and other legal AI tools?
The clearest way to place Spellbook is by what it deliberately does not try to be. Tools like Harvey and Thomson Reuters’ CoCounsel are broader legal-work platforms aimed as much at litigation research, memo drafting, and case-law analysis as at contracts, and are typically sold to larger firms and enterprise legal departments through enterprise agreements. Spellbook stays narrower and cheaper to adopt at the individual-team level: it is a single Word add-in, priced per seat rather than per enterprise deployment, aimed squarely at commercial contract drafting and review rather than case-law research. That makes it a strong fit for a solo transactional attorney, a two-to-twenty-lawyer commercial or corporate practice, or an in-house counsel team at a company under roughly a hundred employees that is drowning in vendor paper — and a weaker fit for a litigation-heavy firm, or any legal department that needs a full contract lifecycle system with a repository and e-signature built in rather than bolted on through a separate tool.
Spellbook is not trying to replace a lawyer’s judgment, and neither Spellbook nor any independent reviewer suggests otherwise: every redline, benchmark, and drafted clause it produces is a first pass that still needs a licensed attorney’s review before it reaches a signature block. What it does well, based on its own published feature set and independent coverage, is remove the blank-page and first-redline problem from high-volume commercial contract work without asking a lawyer to leave Microsoft Word to get there.
Frequently asked
What is Spellbook and how does it work inside Microsoft Word?
Spellbook is an AI contract drafting and review add-in that installs directly into Microsoft Word for Windows, Mac, and the web, appearing as a sidebar next to the document a lawyer is editing. Rather than requiring lawyers to paste contract text into a separate chatbot, it reads the open document and applies suggestions as native Word Track Changes, which a lawyer can accept, reject, or edit like any other redline. Its core tools are named Review, Draft, Ask, Compare, and Playbooks, each targeting a specific step of contract work rather than one general-purpose chat box, which is the main reason transactional lawyers describe it as fitting into an existing workflow rather than replacing it.
What AI models does Spellbook use?
Spellbook was originally built on OpenAI's GPT-4 shortly after its 2023 rebrand from Rally. The company's current site states the product is now powered by state-of-the-art large language models including GPT-5 and Anthropic's Claude (referred to by its Opus model line), meaning requests are routed across more than one provider rather than a single model. Spellbook's terms of service describe a Zero Data Retention arrangement with those underlying model providers, under which contract data sent to a third-party LLM is used only in real time to generate the requested output and is not retained or used for further model training.
How much does Spellbook cost?
Spellbook does not publish per-seat dollar pricing. Its pricing page splits plans into two tracks, one for law firms and one for in-house legal teams, both built on the same underlying Spellbook Suite (the Word add-in plus the Spellbook Associate agent), and both stating that cost is determined by the number of team members on a license. Getting an actual quote requires booking a demo, though a seven-day free trial is available for individuals and teams who want to test the product first. There is no self-serve monthly checkout with listed tier prices, so any specific dollar figure seen on a third-party comparison site should be treated as an unverified estimate rather than Spellbook's published rate.
What is Spellbook Associate?
Spellbook Associate is described by the company as an AI agent built for multi-document transactional work, as opposed to the core Word add-in's one-document focus. Rather than reviewing a single contract at a time, Associate is designed to work across a group of related documents in one project, such as a data room during diligence, a stack of financing documents, a set of disclosure schedules, or a bundle of employment agreements, coordinating analysis and revisions across the whole set. It is bundled together with the Word add-in under what Spellbook calls its 'Spellbook Suite' on both of its published plan tracks rather than sold as a separate add-on.
Is Spellbook secure enough for confidential contract data?
Spellbook states it is SOC 2 Type II audited and supports compliance with GDPR, CCPA, and Canada's PIPEDA. Its terms of service specifically address how contract data is handled by the third-party AI models it uses, describing a Zero Data Retention arrangement in which data sent to an underlying LLM for processing is not retained or used to train that model beyond generating the requested output. As with any AI vendor handling client contract data, firms adopting Spellbook should still review its current data processing agreement and their own jurisdiction's bar guidance on AI confidentiality obligations rather than relying on marketing copy alone.
How is Spellbook different from Harvey or CoCounsel?
Harvey and Thomson Reuters' CoCounsel are broader legal-work platforms that cover litigation research, memo drafting, and case-law analysis alongside contract work, and are typically sold to larger firms and enterprise legal departments. Spellbook stays narrower on purpose: it is a single Microsoft Word add-in focused specifically on commercial contract drafting, review, and benchmarking, priced per team seat rather than through an enterprise-wide deployment. That makes it a common choice for solo transactional attorneys, small commercial or corporate practices, and in-house teams at smaller companies, while larger litigation-heavy firms more often evaluate it alongside, not instead of, a broader research platform.
Who is Spellbook best suited for?
Based on its published feature set and independent reviews, Spellbook fits best for solo transactional attorneys, small-to-midsize firms of roughly two to twenty lawyers doing commercial or corporate work, and in-house counsel teams at companies handling a steady volume of vendor agreements, NDAs, and customer contracts in Microsoft Word. It is a weaker fit for firms that primarily need case-law research, and it does not replace a contract lifecycle management system since it has no built-in repository, e-signature step, or post-signature obligation tracking, so larger legal departments needing full lifecycle coverage typically pair it with, or choose instead, a broader platform.