Contract Review AI Tools Compared: Luminance, Kira, and Spellbook

Scope and Comparison Criteria

This matrix covers three contract review AI platforms that are actively marketed to legal teams and have documented, verifiable feature sets as of Q2 2026: Luminance (Luminance Technologies Ltd.), Kira Systems (now part of Litera), and Spellbook (Rally Legal Inc.). The comparison is scoped to contract review and drafting assistance workflows. It does not cover eDiscovery, legal research, or litigation support use cases, even where a vendor may offer adjacent functionality.

Eight criteria are evaluated. Each is defined below before the comparison table is presented.

• Primary workflow fit — whether the tool is optimized for clause extraction and risk flagging, redlining and negotiation, or AI-assisted drafting from scratch.
• Deployment model — cloud-hosted SaaS, on-premises installation, or hybrid options.
• Underlying AI approach — supervised ML trained on legal corpora, large language model (LLM) integration, or a hybrid architecture.
• Clause extraction coverage — number of out-of-the-box provision types the platform can identify without custom training.
• Drafting / redlining support — whether the tool generates suggested language, redlines against a playbook, or is limited to extraction and flagging.
• Data retention policy — vendor's stated policy on whether uploaded documents and queries are retained for model training or other purposes.
• Jurisdiction coverage — explicitly supported legal systems and languages.
• Pricing structure — publicly disclosed pricing model (per-seat, per-document, enterprise contract, or undisclosed).

Comparison Matrix

Platform Profiles

Luminance

Luminance was built around a supervised ML core trained specifically on legal documents — not general text — which gives its extraction layer a different character from LLM-first tools. The platform's strength is in volume review: large due diligence exercises where a team needs to identify non-standard provisions across hundreds of contracts simultaneously.

The LUMI generative layer, added progressively from 2023 onward, extended the platform into drafting territory. As of Q2 2026, LUMI supports clause-level drafting suggestions and deviation flagging against a firm's fallback positions, though the drafting capability is generally considered secondary to the extraction and analysis features.

Luminance's multilingual support — documented at 70+ languages — makes it a practical option for cross-border M&A teams working with non-English document sets. The on-premises deployment option, available for enterprise clients, addresses data residency requirements that cloud-only tools cannot satisfy.

Kira Systems (Litera)

Kira was acquired by Litera in 2022 and has since been integrated into Litera's broader legal workflow suite. The core Smart Fields architecture — supervised ML trained on M&A and commercial contract corpora — remains the foundation. For teams already using Litera products, the bundled licensing can make Kira's contract review capabilities accessible at lower marginal cost.

The post-acquisition period introduced some uncertainty about product roadmap independence. Teams evaluating Kira should assess whether the Litera integration has affected the standalone contract review experience they may have seen in pre-2022 demos. The generative drafting features added after the acquisition are less mature than Luminance's LUMI layer or Spellbook's core offering.

Kira's documented strength remains M&A due diligence: the Smart Fields library covers the provision types most commonly encountered in deal review, and the platform's workflow for surfacing anomalies across a document set is well-documented in practitioner accounts.

Spellbook

Spellbook occupies a different position in the market. It is built on top of an OpenAI GPT-4-class model with a legal-specific prompt layer and playbook integration, and it operates as a Microsoft Word add-in. The user experience is oriented around in-document drafting: a lawyer working in Word can ask Spellbook to draft a clause, explain a provision in plain language, or suggest a redline against their standard position — without leaving the document.

This makes Spellbook's use case materially different from Luminance and Kira. It is not a bulk extraction tool. It does not produce a heatmap of risk across a 200-document data room. Its value is in the individual contract negotiation and drafting workflow — the attorney reviewing a single agreement and wanting AI-assisted suggestions inline.

The public pricing model with a free trial tier makes Spellbook accessible to smaller firms and solo practitioners in a way that enterprise-contract-only tools are not. The tradeoff is that jurisdiction coverage is primarily US law, and the LLM dependency means outputs carry hallucination risk that requires attorney verification — a limitation that applies to all generative AI drafting tools but is more directly relevant here given the core use case.

Where the Three Tools Diverge Most

The clearest dividing line in this comparison is between extraction-first tools (Luminance, Kira) and drafting-first tools (Spellbook). A legal ops director evaluating these three platforms should define this question first: does the team's primary need involve reviewing large volumes of incoming contracts to identify non-standard provisions, or does it involve producing and negotiating outgoing contracts faster?

The answer shapes nearly every other evaluation criterion. Volume extraction workflows favor Luminance or Kira; individual drafting and negotiation workflows favor Spellbook. Teams with both needs may find themselves evaluating whether a single platform can serve both adequately, or whether a two-tool stack is warranted.

Limitations of This Comparison

• Ironclad is not included in this matrix. It is primarily a contract lifecycle management (CLM) platform with AI review features, rather than a standalone AI review tool. A separate CLM-focused comparison would be the appropriate context for evaluating Ironclad alongside tools like Icertis or ContractPodAi.
• Harvey is not included because its contract review functionality is part of a broader legal AI assistant positioning; a direct comparison against extraction-specific tools would misrepresent its scope.
• Pricing figures for Luminance and Kira are not publicly disclosed and therefore cannot be compared on a per-seat or per-document basis. Spellbook's public pricing tiers should be verified on the vendor's current pricing page before procurement.
• EU AI Act classification of these tools — specifically whether any fall under the high-risk AI system category under Annex III — has not been formally published by any of the three vendors as of Q2 2026. Teams subject to EU AI Act obligations should seek written clarification from vendors on conformity assessment status.

Review Cadence and Verification Status

This matrix is reviewed quarterly. The last verification date is May 2026. Feature sets, data retention policies, and pricing structures in this market change with product updates and corporate events — Kira's integration into Litera being the most recent structural example. Readers relying on specific cells for procurement decisions should confirm values against current vendor documentation.