General AI (ChatGPT, Claude, Gemini) is trained on everything and can handle any topic. Legal AI (Harvey, CoCounsel, Relativity) is built or fine-tuned specifically for legal workflows with connections to verified legal databases. The key difference isn't intelligence — it's data access and verification.

General AI knows what the law says based on its training data. Legal AI knows what the law says based on live connections to Westlaw, LexisNexis, or proprietary legal databases. That distinction is the difference between a research memo with hallucinated citations and one you can file in court.

What General AI Does Well for Lawyers

General-purpose AI — Claude, ChatGPT, Gemini — excels at tasks that require language processing rather than legal database access. Drafting: Claude produces better legal prose than any legal-specific tool because it was trained on more diverse, high-quality writing. Summarization: give Claude a 50-page deposition transcript and it'll produce a clean summary in minutes. Analysis: describe a legal scenario and general AI will identify issues, spot arguments, and suggest strategies based on broad legal knowledge. Client communications: drafting emails, letters, and explanations in plain language. Brainstorming: identifying creative arguments, anticipating opposing positions, and stress-testing theories. General AI handles roughly 60-70% of what lawyers need from AI — everything that doesn't require verified citations from a legal database.

Legal AI's advantage is verified data access. CoCounsel searches Westlaw's 40,000+ databases when you ask a research question — it's not relying on training data from 2024. Harvey's agents are trained on firm-specific precedents and templates. Relativity's TAR models are trained on actual document review decisions from experienced attorneys. Specific capabilities general AI lacks: citation-verified research memos (CoCounsel delivers 95%+ citation accuracy by searching primary sources in real time), jurisdiction-specific analysis with current authorities (legal AI accesses live databases, not static training data), e-discovery document classification at scale (Relativity processes millions of documents with court-accepted protocols), and firm-specific output (Harvey produces documents in your firm's style because it learned your templates). The premium you pay for legal AI — $100-1,200/month vs. $20/month — buys you verification infrastructure that general AI doesn't have.

When General AI Is Enough

For most day-to-day legal work, general AI is sufficient. Claude Pro at $20/month covers: first drafts of briefs, motions, and contracts (you verify citations separately). Research analysis where you provide the sources (paste case text and ask for analysis). Client correspondence and communications. Internal memos and case summaries. Deposition preparation and transcript analysis. Contract review for specific clauses or risks. CLE and professional development (explaining complex legal concepts). The rule of thumb: if the task doesn't require citing specific cases or statutes that you haven't already identified, general AI handles it. A solo practitioner doing 80% drafting and 20% research gets 90% of their AI value from Claude Pro. Adding CoCounsel for the research component is a luxury, not a necessity, if they verify citations through Westlaw or Google Scholar manually.

Legal-specific AI becomes necessary in four scenarios. 1. High-volume citation-dependent research: if you're producing 10+ research memos per week, manual citation verification is a bottleneck. CoCounsel eliminates it. 2. E-discovery at scale: no general AI handles TAR/predictive coding workflows. Relativity, Everlaw, and Reveal are purpose-built for this. 3. Enterprise standardization: when 200+ attorneys need consistent output that follows firm-specific templates and style guides, Harvey's custom agents deliver consistency that general AI can't match. 4. Regulatory compliance workflows: tools like Kira Systems and Luminance are trained specifically on contract analysis patterns that general AI approximates but doesn't nail with sufficient precision for M&A due diligence. The deciding factor is risk tolerance and volume. Low-volume, attorney-verified work = general AI is fine. High-volume, production-critical work = legal-specific AI pays for itself.

The Hybrid Approach: Most Firms Use Both

Smart firms don't choose between general and legal AI — they use both for different tasks. Common hybrid stack: Claude (general) for drafting + CoCounsel (legal) for research. Claude handles the 60-70% of work that's language-based. CoCounsel handles the 30-40% that requires verified legal data. The cost: $170-225/user/month vs. $1,200+ for Harvey alone. This hybrid outperforms an all-general or all-legal approach. All-general means every citation needs manual verification (time-consuming). All-legal means overpaying for drafting tasks that don't need database access. The hybrid matches the right tool to the right task. For managing partners: the question isn't "general or legal AI" — it's "which tasks need verified legal data and which don't?" The answer determines your stack composition and your budget.

The Bottom Line: General AI (Claude, ChatGPT) handles drafting, analysis, and communications — 60-70% of legal AI needs at $20/month. Legal-specific AI (CoCounsel, Harvey) adds verified legal databases, citation accuracy, and firm-specific training at $100-1,200/month. Most firms need both. Use general AI for language tasks, legal AI for citation-dependent research and enterprise-scale operations.

AI-Assisted Research. This piece was researched and written with AI assistance, reviewed and edited by Manu Ayala. For deeper takes and the perspective behind the research, follow me on LinkedIn or email me directly.