How Anthropic uses Claude in Legal

Key Concepts

• Claude Code: An AI tool enabling users without coding experience to automate tasks through natural language instructions.
• Workflow Automation: Utilizing Claude to streamline repetitive legal tasks, such as content review and conflict of interest checks.
• Hallucination (AI): The tendency of AI systems to generate incorrect or nonsensical information.
• Risk Level Signal: A prioritization system (Low, Medium, High) implemented within the workflow, based on a user-defined framework, to highlight potential legal issues.
• Redlining: The process of marking changes to a document, often used in contract review.

Automating Legal Workflows with Claude Code

The speaker details their experience leveraging Claude Code, an AI tool, to automate various legal tasks, despite having no prior coding experience. The initial impetus was a personal project – creating a “legal lamp” that blinks messages in Morse code, demonstrating Claude’s ability to translate natural language requests into actionable outputs. This project served as a proof of concept for broader applications within their legal department.

Addressing Repetitive Legal Tasks

A core problem identified was the burden of repetitive, time-consuming “busywork” that diverted legal professionals from higher-level strategic thinking. Before implementing Claude, the speaker frequently faced last-minute requests from the marketing team to review blog posts for potential legal issues, often with tight deadlines. This process was described as “dull” and prone to errors.

The Marketing Materials Review Workflow – A Case Study

The speaker outlines a specific workflow built with Claude to address the marketing review process. This workflow is accessible via a pinned link in their channel ("Marketing Materials Software review tool"). The process is as follows:

1. Content Input: A marketer copies and pastes the blog post content into the review tool.
2. Analysis Initiation: Clicking the “Analyze Content” button triggers Claude to review the material.
3. Issue Identification: Claude identifies potential legal issues, categorizing them based on areas like accuracy, security claims, publicity rights, and partnership considerations. In the example provided, Claude flagged five issues.
4. Positive Feedback: Claude also highlights aspects of the content that are well-written and legally sound, such as clear feature descriptions and comprehensive integration details.
5. Ticket Generation: The user can automatically generate a Slack message summarizing the issues for the legal team and create a legal ticket with prioritized issues (Low, Medium, High risk) for review. The risk level prioritization is based on a framework defined by the user.

Prioritization and Human Oversight

The speaker emphasizes that Claude acts as a “first pass” for identifying potential issues, prioritizing them based on a pre-defined risk framework. However, crucial to the implementation is maintaining “a human in the loop.” This is due to the acknowledged potential for AI “hallucinations” – the generation of inaccurate or misleading information. The speaker continues to review Claude’s output to ensure accuracy and prevent errors. As stated, “I’m still making sure I’m reviewing the work.”

Expanding Applications of Claude within the Legal Department

Beyond marketing materials review, Claude is being utilized in several other areas:

• Redlining: Assisting with the review and marking of changes in commercial contracts.
• Conflict of Interest Policy: Supporting the review of potential conflicts of interest.
• Outside Business Activity Requests: Evaluating requests from employees regarding external business ventures.

Getting Started and Key Takeaway

The speaker advises those interested in utilizing Claude to begin by identifying their most routine and repetitive tasks. They encourage experimentation, stating, “You really don’t know what it’s capable of doing until you give it a shot.” The overall benefit highlighted is increased speed and proactive issue identification within the legal department, allowing legal professionals to focus on more strategic work.

Notable Quote: “No lawyer likes doing the same repetitive exercise over and over and over again. The work can be dull. You can make mistakes.” – Speaker, describing the motivation for exploring AI automation.