Sullivan & Cromwell brings legal expertise into agreement review
OpenAI and Sullivan & Cromwell introduced Agreement Analyzer, an AI tool that reviews draft agreements by comparing them to precedents and playbooks, producing issue reports and proposed changes for lawyers to refine.
Sullivan & Cromwell and OpenAI have jointly developed Agreement Analyzer, an AI-powered tool designed to assist lawyers in reviewing draft agreements. The tool integrates the firm’s precedents, negotiating playbooks, and ChatGPT Enterprise to analyze agreements against established standards. Lawyers can input a draft, select relevant precedents and playbooks, and receive a synthesized review highlighting potential issues and proposed changes. The output includes a memorandum and tracked changes for further review or adjustment.
Agreement Analyzer was initially created by Sullivan & Cromwell partner John Savva as a single-user proof of concept using Codex. After collaboration with OpenAI, the tool was enhanced to support multi-user workflows, incorporating parallel processing for efficiency. The application now allows multiple lawyers to use the tool simultaneously while maintaining transparency in its analytical process. Savva notes that the tool’s adversarial review system helps validate findings by challenging conclusions before finalizing results.
The tool performs multiple review cycles from different perspectives, identifying issues such as market deviations, risk allocation conflicts, clause interactions, and ambiguities that could impact litigation. Findings are categorized as supported, dismissed, or unverified, with explanations and precedents provided for each. Lawyers can inspect the issues, their supporting evidence, and decide which findings to incorporate into the final agreement markup.
Sullivan & Cromwell is continuing to explore AI tools tailored to specific legal tasks, with Savva highlighting advances in reasoning capabilities with GPT-6 Astra. He states that Astra has improved the quality of reasoning, addressing a previous bottleneck in AI-assisted legal analysis. The firm aims to expand the use of AI to support more of the research and writing process in legal practice.