Spellbook

Spellbook: What Sits Inside Word

• Works where lawyers work — Word and Google Docs, not a separate application.
• Redlines against your standards, using playbooks rather than generic best practice.
• Zero data retention: model providers stated never to train on, learn from or store your data.
• Market comparison against thousands of similar agreements.
• SOC 2 Type II, HIPAA, GDPR and the EU AI Act listed, across 80-plus countries.

Spellbook: Coverage And Compliance

Spellbook: Redlines Against Your Own Standards

The proposition is contracts at the speed of commerce: review contracts, draft with your standards, and search every deal you have signed.

The three-stage framing underneath it — intake, review, insight — describes a lifecycle rather than a feature, and that is the more useful way to read the product.

The phrase doing the real work is your standards. A tool that reviews a contract against general best practice is telling you what a textbook thinks, and lawyers already know that.

What a firm actually needs is whether this agreement departs from the position it has taken on the last four hundred, because a term that is unremarkable in the abstract may be one the firm has decided never to accept.

That is what playbooks encode, and it is why the same clause can be fine at one organisation and an automatic escalation at another. The value is not legal knowledge; it is institutional memory applied consistently at three in the morning on the fortieth vendor agreement of the quarter.

A tool reviewing against general best practice tells lawyers what a textbook thinks. What a firm needs is whether this departs from the position it took on the last four hundred.

Playbooks And What Non-Standard Means

The homepage carousel puts it plainly: catches non-standard terms across vendor agreements.

Non-standard is a relative word, and encoding what it means for a specific organisation is where this category succeeds or fails.

Every legal team already has a playbook, even when it is not written down. It lives in a partner's head, in a folder of precedents, and in the reflex of an associate who knows the firm never accepts unlimited indemnity in this deal type. New joiners acquire it by getting things sent back.

Writing that down is valuable independent of any software, and it is the part of adoption most likely to be underestimated.

The corollary is that output quality tracks the effort put into the playbook. A team that configures carefully gets review reflecting how it actually negotiates; a team that installs and expects magic gets generic commentary and concludes the category does not work. That is a deployment problem rather than a product one, and it is the same story with enterprise platforms like Ironclad at the contract management end of the same lifecycle.

Why Does Contract Review Break At Volume?

The failure is not that lawyers cannot review contracts. It is that the volume of routine agreements has grown faster than the capacity to look at them properly.

An in-house team of four receives vendor agreements, NDAs, order forms and renewals from every department, arriving through email, Slack and the CRM. Most are unremarkable. A small number contain something that matters.

The economics force a choice nobody likes: review everything slowly and become the department that blocks deals, or triage by instinct and accept that something will eventually be signed unread.

A first pass on every contract changes that arithmetic, because the question shifts from which agreements to read to which flagged terms to consider.

The site's own quotes describe both sides of the benefit — a partner estimating an extra billable hour a day, and a lawyer at a technology company describing the substantive work being run through it. The second framing is the more interesting one: not doing less legal work, but spending the same hours on the parts that need judgement.

• Write the playbook down — it is valuable with or without software.
• Confirm zero retention terms with your own counsel, not from a homepage.
• Start on one high-volume agreement type, not the whole intake.
• Measure what the first pass catches that a human missed, and vice versa.
• Keep the lawyer accountable for every position that ships.

Zero Retention And Legal Confidentiality

The security statement is unusually direct: built for legal-grade confidentiality, with zero data retention meaning model providers never learn from, train on or store your data, alongside SSO, MFA, encryption and audit controls.

For this profession that is not a feature, it is the entry requirement.

Legal privilege and professional confidentiality obligations are not satisfied by a vendor's good intentions. A firm sending client documents to a third-party model has to be able to say precisely where they went, how long they persisted and who could see them — and for many firms the honest answer historically was that nobody knew.

Naming the commitment at provider level rather than only at vendor level is the specific detail that matters, because that is where the uncertainty usually sits.

The compliance list is broader than the norm: SOC 2 Type II, HIPAA, GDPR and the EU AI Act. That last one is worth noticing, because the EU AI Act imposes obligations by risk category and few vendors in any sector list it yet. Whether that reflects completed assessment or stated intent is a fair procurement question, and the answer belongs in writing. Firms comparing this against broader platforms like Harvey AI should compare the retention terms line by line rather than the feature lists.

Searching Every Deal You Have Signed

The third stage is the one that quietly justifies the platform: signed contracts stored automatically and indexed so you can search your full history, plus market comparison against thousands of similar agreements.

Most organisations cannot answer basic questions about their own contracts.

Which of our agreements auto-renew next quarter? How many contain a most-favoured-nation clause? What did we agree on liability caps with customers in this segment? These are answerable in principle and, in practice, require a person opening files for a week — so they are asked once, answered approximately, and not asked again.

An indexed repository turns them into queries, and the value compounds because every signed agreement makes the next answer better.

Market comparison is the more novel half, and it is worth being precise about what it is. Knowing that a term appears in most similar agreements tells you what is customary, not what is acceptable to you — those are different, and confusing them is how a negotiating position erodes one standard clause at a time. Used as context it is genuinely useful; used as an argument it is how everyone converges on the same terms. Document review platforms such as Luminance approach the corpus problem from the diligence side rather than the drafting side.

Reviewing A Contract With AI Assistance

The risk in legal AI is not a wrong answer. It is a plausible answer that a busy person accepts without checking.

This order keeps the lawyer where the accountability sits.

1. Have your own counsel review the retention and confidentiality terms first.
2. Write the playbook down before configuring anything.
3. Pilot on one high-volume, low-risk agreement type.
4. Compare the first pass against a human review of the same documents.
5. Treat market comparison as context, never as a negotiating position.
6. Keep a named lawyer accountable for every term that gets signed.