2 / 26
AI Fundamentals for Legal Professionals· 18 min read·Personalise for your role →

Choosing the Right AI Tools for Legal

Navigate the growing landscape of legal AI tools — from general-purpose models to specialist platforms — and understand what questions to ask before adopting any tool.

In practice: Contract first-pass: 2–4 hours → 20 minutes

Step 1 of 4

Read

0% done

Two Categories of Legal AI

General-purpose LLMs (Claude, ChatGPT, Gemini) excel at drafting, summarisation, and analysis. They are flexible and powerful but have no built-in legal database access.

Legal-specific platforms (Harvey, Clio Duo, Westlaw AI, LexisNexis+AI) combine LLM capability with legal database integration. They offer citation checking, jurisdiction awareness, and practice management integration.

What to Evaluate Before Adopting Any Tool

Data Governance

  • Where is data processed and stored?
  • Is your input used to train the model?
  • Are there enterprise data protection agreements available?
  • What certifications does the vendor hold (SOC 2, ISO 27001)?

Accuracy and Hallucination Risk

  • Does the tool cite its sources?
  • Can you verify outputs against authoritative databases?
  • How does the vendor handle known hallucination risks?

Integration

  • Does it connect to your matter management system?
  • Can it access documents from your DMS without manual uploads?
  • How does it handle billing and time recording?

The General-Purpose Workflow

For most legal teams without a specialist platform, a practical workflow is:

  1. Use Claude or ChatGPT for drafting, analysis, and summarisation
  2. Verify all legal references in Westlaw, LexisNexis, or official government sources
  3. Apply your professional judgment to AI outputs before any client use

What to Avoid

  • Using free-tier tools for client-related work (data protection issues)
  • Treating AI-generated research as a substitute for verified primary sources
  • Adopting a tool before getting IT and risk management sign-off
Tool evaluation: surface-level vs. governance-focused assessment

The prompt

Compare ChatGPT and Harvey for legal work

Claude's response

ChatGPT and Harvey are both AI tools used in the legal profession. ChatGPT is a general-purpose tool made by OpenAI that can help with many tasks. Harvey is a legal-specific AI platform designed for law firms. Both can assist with drafting and research. Harvey is more specialised for legal work. ChatGPT is more widely available. The choice depends on your firm's needs and budget.

Quick check

Your firm is evaluating a free-tier AI chatbot for legal research. A partner argues it is fine because the tool produces accurate-looking summaries. What is the critical problem with this approach?

Key Takeaways

  • 1General-purpose LLMs and legal-specific platforms serve different use cases
  • 2Evaluate any tool on data governance, accuracy, and integration before adoption
  • 3Enterprise agreements with appropriate data protections are required for client work
  • 4Specialist legal AI adds citation verification that general-purpose tools cannot provide
  • 5Always verify legal references through authoritative databases regardless of which AI tool you use

Your challenge this week

Apply what you learned in a real task

Check whether the AI tool you currently use (or plan to use) for any legal work has an enterprise data processing agreement in place with your organisation. If not, identify what steps are needed before you can use it for client-related work.

Starter prompt · paste into Claude or ChatGPT
I am evaluating [AI tool name] for use in a [type of legal practice/organisation]. Help me draft a list of data governance questions I should ask the vendor before approving use for client matters. Focus on: data retention, training use, data residency, breach notification, and any relevant [UK/EU/other] regulatory requirements.

Before you practise

What AI tools are currently being used — formally or informally — in your team or organisation? Are you confident that each of those tools has the data governance controls required for the type of content being submitted? Where are the gaps?

Was this lesson helpful?

Next step

Put it into practice

You've read the lesson — now apply it in a guided hands-on exercise. It takes about 5 minutes.

Ask the AI Tutor