Sullivan & Cromwell Integrates AI-Powered Agreement Review Tool

Sullivan & Cromwell, a leading international law firm, has developed an innovative tool to enhance the review process of draft agreements. The firm collaborated with OpenAI to create Agreement Analyzer, an artificial intelligence-powered tool designed to identify potential issues in a draft agreement.

The development of Agreement Analyzer was led by John Savva, a partner at Sullivan & Cromwell. According to Savva, AI has evolved significantly over time and is now being used for more complex tasks such as deep analysis of agreements. This shift allows lawyers to focus on high-level decision-making rather than repetitive administrative tasks.

Agreement Analyzer integrates various components, including the firm’s curated precedents, negotiating playbooks, and ChatGPT Enterprise. The tool enables users to input a draft agreement and add selected precedents, along with information on market practice, client positions, and firm positions. It then performs multiple review cycles from different perspectives, synthesizing outputs and identifying potential issues.

The output of Agreement Analyzer includes a memorandum and proposed tracked changes for the lawyer’s review, adaptation, or rejection. Savva emphasizes that AI tools should provide transparency in their decision-making processes. He notes that users need to be able to see the source of information used by the tool when making claims about market practice.

Savva also discussed potential use cases where specialized agents review agreements in parallel for specific aspects such as market deviations, risk allocation, and ambiguities. The results distinguish between findings supported after challenge, dismissed findings, and unverified issues retained with qualifications. Lawyers can inspect these issues along with their supporting explanations and relevant precedents.

Sullivan & Cromwell is continuing to explore the development of AI-powered tools tailored to specific legal tasks. Savva believes that advances in reasoning capabilities, such as those seen in GPT-6 Astra, have created opportunities for more comprehensive support of research and writing processes.