Drafting legal documents is where AI should shine. The structure is often formulaic, the language is precedent-driven, and the workflow is repetitive. A motion to dismiss follows the same structure every time. A commercial lease has the same 40 clauses with different values. An engagement letter is 90% boilerplate.

So why do most AI drafting tools produce output that takes longer to edit than writing from scratch?

The problem is specificity. Legal drafting requires jurisdiction-specific language, client-specific terms, and matter-specific arguments. Generic AI produces generic output. The tools that work are the ones that let you control the inputs precisely enough to get usable first drafts.

What I Tested

I tested 6 AI drafting tools by asking each to produce three documents:

1. A motion to dismiss for lack of personal jurisdiction (federal court)
2. A commercial software licensing agreement (B2B SaaS)
3. A client engagement letter (litigation matter)

I evaluated each tool on accuracy, customization, time saved, and edit burden (how much work the draft needed before it was usable).

The Tools

1. Spellbook

Spellbook lives inside Microsoft Word and drafts documents based on your instructions and existing templates.

How it works: You open Word, describe what you need, and Spellbook generates a draft in your document. It learns from your firm’s existing documents to match your style and terminology.

Results:

• Motion to dismiss: Produced a solid structure with proper caption, standard arguments, and correct formatting. Needed 45 minutes of editing to add case-specific arguments and jurisdiction-specific language.
• Software license: Good boilerplate, but missed some SaaS-specific provisions (data processing terms, SLA uptime requirements). Needed 1.5 hours of editing.
• Engagement letter: Best result. Produced a clean, professional letter that needed only 20 minutes of customization.

Edit burden: Medium. Good structure, needs case-specific substance.

Pricing: $199/month per user

Best for: Firms that draft in Word and want AI integrated directly into their existing workflow.

2. Harvey AI

Harvey drafts documents through its chat interface, using your firm’s knowledge base and legal research capabilities.

How it works: You describe the document you need, provide context (case facts, client information, jurisdiction), and Harvey generates a draft with citations to supporting law.

Results:

• Motion to dismiss: Strongest output of any tool. Included relevant case citations, proper legal standards, and persuasive argument structure. Needed 30 minutes of editing to add specific facts and adjust tone.
• Software license: Decent but generic. Missed industry-specific terms. Needed 2 hours of editing.
• Engagement letter: Good but overly detailed. Needed 30 minutes of trimming and client-specific additions.

Edit burden: Low for litigation drafting, medium to high for transactional documents.

Pricing: $50-100/user/month

Best for: Litigation practices that need AI-researched first drafts for motions and briefs.

3. Gavel (formerly Documate)

Gavel is a document automation platform that uses conditional logic to generate legal documents from templates.

How it works: You build or import templates with conditional logic (if client is married, include spousal provisions), then run an interview to populate the template. It’s not generative AI in the ChatGPT sense; it’s smart template automation.

Results:

• Motion to dismiss: Only works if you build the template first. No first-draft capability from scratch.
• Software license: Same. Template-based, not generative.
• Engagement letter: Excellent if you have a template. The interview process ensures all necessary information is collected.

Edit burden: Very low (if template is well-built). Very high (if starting from scratch, because you have to build the template first).

Pricing: $99-299/month depending on features

Best for: Firms that produce high volumes of similar documents (estate planning, immigration, real estate) and are willing to invest in template creation.

4. LawDroid

LawDroid is a legal AI platform that includes document drafting, client intake, and chatbot capabilities.

How it works: You use their builder to create document workflows. Clients or staff answer questions, and the system generates documents based on the answers.

Results:

• Motion to dismiss: Not designed for this type of drafting. Output was unusable.
• Software license: Produced a basic template that needed significant customization. 3+ hours of editing.
• Engagement letter: Similar to software license. Too generic for practical use without heavy editing.

Edit burden: High for complex documents, low for simple client-facing documents.

Pricing: $29-99/month

Best for: Solo practitioners who want an affordable client-facing chatbot that can also generate simple documents.

5. EvenUp

EvenUp focuses specifically on personal injury demand letters and case preparation documents.

How it works: You upload medical records, police reports, and case files. EvenUp’s AI generates a demand letter with damage calculations, liability analysis, and settlement demand.

Results:

• Motion to dismiss: Not applicable (wrong practice area).
• Software license: Not applicable.
• Engagement letter: Not applicable.
• Demand letter (bonus test): Exceptional. Produced a detailed, well-organized demand letter that needed minimal editing. Saved 4-6 hours per demand.

Edit burden: Very low for its specific use case.

Pricing: Per-document pricing, typically $500-1,000 per demand package

Best for: Personal injury firms that produce high volumes of demand letters.

6. Local AI (Ollama + Open Models)

Running a local AI model gives you complete control over the drafting process with no data leaving your office.

How it works: You install Ollama, download a model like Qwen 2.5 14B, and use it to draft documents in your terminal or through a local interface.

Results:

• Motion to dismiss: Produced a usable structure with standard arguments. Needed 1 hour of editing to add case law and specific facts. Quality depends heavily on the model and prompt.
• Software license: Decent boilerplate but missed key provisions. 2 hours of editing needed.
• Engagement letter: Good quality. 30 minutes of editing.

Edit burden: Medium. Output is decent but requires more guidance than specialized legal AI tools.

Pricing: Free (requires hardware with 16-32GB RAM)

Best for: Privacy-conscious firms handling privileged documents. See our complete local AI setup guide for lawyers.

Which Tool Should You Choose?

For litigation drafting (motions, briefs, memos): Harvey AI. Its research integration means the first draft comes with supporting citations, saving you the most time.

For transactional drafting (contracts, agreements): Spellbook if you want Word integration, Gavel if you have high volume and can invest in templates.

For client communications (engagement letters, status updates): Spellbook. The Word integration makes it easy to customize output, and engagement letters are its strongest use case.

For personal injury demand letters: EvenUp. Nothing else comes close for this specific use case.

For privacy-sensitive drafting: Local AI using Ollama. The quality is lower than specialized tools, but client data never leaves your office.

For budget-conscious solo practitioners: LawDroid for simple documents, supplemented by local AI for more complex drafting.

How to Get Better Output From Any AI Drafting Tool

Regardless of which tool you use, these practices improve output quality:

• Provide specific context: “Write a motion to dismiss for lack of personal jurisdiction in a breach of contract case where the defendant is a Delaware LLC with its principal place of business in Texas, the contract was executed in New York, and the only alleged contact with this jurisdiction is a single email” produces dramatically better output than “Write a motion to dismiss for lack of personal jurisdiction”
• Include your jurisdiction: State the court, the circuit, the applicable law. AI tools produce better output with jurisdictional context
• Specify your audience: “This will be filed in federal court” versus “This is an internal memo” produces different levels of formality and citation density
• Feed it examples: If you have a previous motion that won, paste it as a reference. “Match the style and structure of this document” works well
• Edit in stages: Get the structure right first, then fill in substance, then polish language. Trying to perfect everything in one edit pass is slower

The Bottom Line

AI drafting tools save 1-3 hours per document when used correctly. The key is matching the tool to your practice area and providing enough context to get usable output. No tool produces court-ready documents without editing, but the right tool reduces editing from “rewrite the whole thing” to “fill in the blanks and verify citations.”

Related reading: AI Legal Document Drafting for Lawyers · Spellbook Review for Lawyers · Best AI Tools for Lawyers · AI Contract Review Tools for Lawyers · Local AI for Lawyers: Private, Free · AI for Small Law Firms on a Budget · AI Legal Research Prompts for Lawyers

Not a lawyer? See AI tools for teachers or best AI tools by profession.

FAQ

Can I file AI-drafted documents with a court?

Yes, but you must review and verify the output before filing. Several courts now require disclosure of AI use in document preparation. Check your local rules. The document must reflect your professional judgment, not just AI output.

Is AI drafting good enough for client-facing documents?

For templates and standard documents (engagement letters, NDAs, simple agreements), yes. For complex, high-stakes documents (M&A agreements, sophisticated financing documents), AI produces a useful first draft but requires substantial attorney input on business terms and risk allocation.

How do I prevent AI from hallucinating case citations in drafts?

Provide the case citations yourself or use tools like Harvey that research real cases. Never trust AI-generated citations without verification. Better yet, draft the argument yourself and use AI to polish the language and structure.

What’s the best AI tool for solo practitioners on a budget?

Local AI (free, requires hardware investment) or LawDroid ($29/month) for simple documents. For more capable tools, Spellbook at $199/month is the best value for solo practitioners who draft frequently.

How do I train AI to match my firm’s writing style?

Tools like Spellbook and Gavel let you upload existing documents as examples. The AI learns your preferred terminology, clause structure, and formatting. For tools without this feature, paste an example document as a reference in your prompt and explicitly instruct the AI to match the style.

Related Articles

• AI for Lawyers: Running Models Locally
• AI Legal Research Prompts for Lawyers 2026
• Best AI for Case Law Research 2026
• AI Deposition Summary Tools
• Local AI Models for Law Firms
• AI Tools for Teachers