Counsel Ops HQLitigation Matter & Spend Tech
Comparison

Spellbook vs Luminance vs LegalOn: Which AI Contract Tool Belongs in Your Word Toolbar?

Spellbook, Luminance and LegalOn all work inside Microsoft Word, but they differ on playbooks, jurisdictions, data terms and pricing. Here's which one fits a solo practice, a mid-size firm or an in-house team.

By the Counsel Ops HQ editorial team · November 2026 · 9 min read

Three Word Add-ins, Three Different Buyers in Mind

If you're choosing an AI contract tool this budget cycle, your shortlist probably includes Spellbook, Luminance and LegalOn. All three run inside Microsoft Word, all three suggest redlines, and all three say they can apply your positions to the other side's paper. The similarity ends there. Each was built for a different buyer, and if you choose on demo polish alone, you can end up with a tool your lawyers stop opening after six weeks.

The short version. Spellbook, from Canada, is a drafting-first assistant for transactional lawyers who want speed across a wide mix of agreements. Luminance, from the UK, grew out of due diligence and now positions itself as an enterprise platform built around its own legal-trained models, with negotiation features aimed at first-pass redlines on standard agreements. LegalOn, which started in Japan and later launched an English-language product, is review-first. Its main selling point is a library of attorney-built playbooks you can use from day one.

Advertisement

That's why a Spellbook vs Luminance decision is rarely a straight feature comparison. It's a choice between a flexible drafting partner and a governed, organization-wide system. LegalOn sits between them: more structured than Spellbook, lighter to deploy than Luminance. Five criteria separate the three: how playbooks get built, what stops a bad redline from going out, which governing laws and languages each handles well, what happens to your data, and how seats are priced.

This comparison draws on each vendor's public positioning and on how tools like these are usually deployed. We haven't run hands-on tests, so treat it as a framework for your own pilot, not as results.

Playbook Setup and the Controls That Catch Bad Redlines

Most of a contract tool's value comes from the playbook: your preferred positions, fallbacks and walk-away points. The effort to build one differs sharply across these three.

LegalOn: fastest to a usable playbook

LegalOn ships with prebuilt playbooks for common agreements such as NDAs, MSAs, SaaS terms and data processing agreements. You adjust them to your risk appetite instead of starting from a blank page. For an in-house team that sees the same handful of contract types every month, this is the biggest time-saver of the three. The trade-off: the defaults reflect the vendor's idea of a reasonable stance. If your positions are unusual, you'll still be editing.

Luminance: most governed, most setup effort

Luminance is built to learn from your historical contracts and the standards you configure. For a large team with a deep archive, that pays off, because deviations are flagged against what your organization actually signs, not a generic market norm. Expect a real implementation project, though: gathering and cleaning precedent documents, assigning owners for clause positions, and usually some vendor-led onboarding. Plan for anywhere from several weeks to a few months before redlines are consistently trustworthy, depending on how tidy your precedent bank is.

Spellbook: lightest playbooks, most drafting freedom

Spellbook lets lawyers set review guidelines and reuse them, but its real strength is generating and revising clauses on request, plus workflows that take a multi-step drafting instruction and work through a document. That suits a lawyer who knows what they want and wants it typed faster. It does less to enforce consistency across 30 lawyers with different habits.

Quality controls to test in any pilot

Check these before signing:

  • Tracked changes, never silent edits. Every suggestion should appear as a standard Word tracked change with an explanation the reviewer can read.
  • Source attribution. Can the tool show which playbook rule or precedent triggered a redline? An unexplained suggestion is a liability.
  • Defined-term integrity. Feed it a 40-page MSA with nested definitions and see whether it breaks cross-references or invents defined terms.
  • Approval gates. In-house teams should confirm that positions outside the playbook can be routed for sign-off rather than accepted by a junior reviewer.
  • Audit trail. Can you later see who accepted which AI-generated change, and when?

Treat any vendor that measures redlining quality only by speed as a red flag. Ask what share of suggestions your reviewers rejected during the pilot, and why.

Governing Law, Language Coverage and Data Retention Terms

This is where a global readership needs to slow down. All three tools rely on large language models, and how well they handle a contract depends on how much training material and playbook content exists for that jurisdiction and language.

Governing law. Spellbook and LegalOn's English-language product are most at home with common-law commercial contracts, the kind governed by the laws of England and Wales, New York, Delaware, Ontario, Singapore or Australia. LegalOn's original product was built around Japanese law and Japanese-language contracts, which matters to a team in Tokyo or a regional group with Japanese counterparties. Confirm which country editions you can actually buy, because scope and availability differ by market. Luminance markets itself as language- and jurisdiction-agnostic and sells internationally, including in Europe, the Middle East and Asia-Pacific. That makes it the safer default for a group negotiating under German, French, UAE or Brazilian law in the same quarter, subject to your own testing.

Language. If a meaningful share of your contracts aren't in English, test each tool on your own documents. Don't rely on a vendor's language list. Clause recognition in Spanish, Arabic or Bahasa can trail English. Civil-law concepts such as mandatory statutory warranties, notarial formalities or local-language precedence clauses are where generic suggestions most often go wrong.

Data retention and training. Each vendor's public materials address whether customer documents are used to train shared models, and each describes enterprise security controls and certifications. Verify the current position and get it into the contract, not just a sales deck. Ask for:

  • A data processing agreement that names every subprocessor, including the underlying model providers.
  • Written zero-retention or short-retention terms for prompts and outputs sent to third-party models.
  • Hosting region options. Clients under the EU or UK GDPR, teams subject to India's Digital Personal Data Protection Act, and Gulf clients facing sector-specific localization rules may each need a specific region.
  • Deletion timelines when the contract ends, and a certificate of deletion.
  • Confirmation that client-confidential material may be processed at all under your engagement letters and your bar or law society's guidance on generative tools.

Professional-conduct rules on confidentiality and technology supervision differ by jurisdiction, as do data transfer rules. Have your data protection officer or outside privacy counsel review the vendor terms before rollout.

Seat Pricing and What Actually Drives the Bill

None of the three publishes a full enterprise price list, and quotes vary by seat count, contract length, region and bundled modules. The ranges below are typical 2026 market ranges based on how these tools are commonly sold. They're indicative, in USD, and will vary by contract and country. Confirm current pricing with each vendor.

FactorSpellbookLuminanceLegalOn
Typical pricing modelPer seat, monthly or annualAnnual enterprise license, often platform-basedPer seat, annual
Indicative costRoughly $100–$500 per user per monthRarely sold per seat to small teams; total contracts commonly in the tens of thousands to well over $100,000 per yearRoughly $100–$400 per user per month
Minimum commitmentLow; small teams and solos can buyHigh; usually enterprise minimumsModerate; team plans
Implementation effortDaysWeeks to monthsTypically a few weeks
Best fitSolo to mid-size transactional practicesLarge firms and multinational legal departmentsIn-house teams and mid-size firms with repeat contract types

What moves the number:

  • Seat definitions. Find out whether occasional reviewers, paralegals and contract managers need full seats or can have cheaper access.
  • Module bundling. Luminance in particular sells negotiation, repository and diligence capabilities that may or may not be in your quote.
  • Term length. Multi-year commitments often earn a discount, but they lock you in while the category is still changing quickly.
  • Hosting and security add-ons. Dedicated regions, single sign-on and custom retention terms can carry surcharges.

Hypothetical worked example: Say a 12-lawyer in-house team in Singapore needs eight active reviewers. At $250 per seat per month, LegalOn or Spellbook would cost about $24,000 a year. If each reviewer saves two hours a week over roughly 46 working weeks, at an internal cost of $120 an hour, that's about $88,000 of capacity. Even if real savings come in at a third of that estimate, around $29,000, the tool still covers its cost. Luminance's quote for the same team could be higher, and it wins on value only if the team also needs multi-jurisdiction coverage or deeper governance.

The Verdict: Which Tool Fits Which Buyer

Solo and small-practice commercial lawyers: Spellbook. It's the easiest to buy without an enterprise contract, the quickest to start using, and the strongest of the three at drafting from scratch, which is most of a solo commercial lawyer's day. A team of one doesn't need a governance layer. It needs faster first drafts and a second pair of eyes on incoming paper. If you mostly work on English-language, common-law contracts, this is the clear pick.

Mid-size firms (roughly 20 to 150 lawyers): it depends on your workload. Where partners want associates drafting faster across varied deal work, Spellbook usually fits better; pilot it with one practice group first. Where the firm handles high-volume, repeatable client contracts, such as outsourced commercial review for corporate clients, LegalOn's ready-made playbooks give more consistent output across associates with less setup. Firms with heavy cross-border, non-English work should get a Luminance quote too, even though it will cost more.

In-house legal teams: LegalOn for most, Luminance for multinationals. Most corporate legal departments review the same few agreement types repeatedly and care more about consistent positions than creative drafting. LegalOn's preloaded playbooks and review-first design match that pattern and should be useful within weeks. If you're a multinational negotiating in several languages and governing laws, or you want negotiation on standard NDAs largely automated, Luminance's heavier implementation is justified. On a Spellbook vs Luminance shortlist for a global legal department, Luminance usually wins on coverage and control. Spellbook wins only if the main pain point is drafting speed rather than consistency.

Before you sign, run a two-week pilot with 20 of your own recently negotiated contracts, including at least five under a non-English governing law if that applies to you. Track how many suggested redlines each reviewer accepts, edits or rejects. Choose the tool with the lowest rejection rate on your highest-volume contract type, provided its data processing agreement passes your privacy review.

Questions buyers ask

Is Spellbook or Luminance better for contract review in Microsoft Word?

Both work as Word add-ins, but they're built for different jobs. Spellbook is stronger for drafting and fast review by individual lawyers, while Luminance is better for organization-wide, governed review across many languages and jurisdictions. Small practices usually prefer Spellbook. Large multinational legal teams usually prefer Luminance.

Do these AI contract tools train their models on my client documents?

Each vendor's public materials address this, but marketing statements aren't contractual commitments. Ask for a data processing agreement that lists every subprocessor, sets retention periods for prompts and outputs, and offers your required hosting region. Have your privacy counsel review those terms, because data transfer and confidentiality rules differ by jurisdiction.

Can LegalOn contract review handle contracts governed by non-English law?

LegalOn's English-language product is strongest on common-law commercial agreements, and its original product was built for Japanese law and Japanese-language contracts. Coverage of other civil-law systems and languages may be more limited, and availability varies by market. If you regularly negotiate under other governing laws, test it on your own documents during a pilot and compare it against a multilingual platform such as Luminance.

Written by the Counsel Ops HQ editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.