Skip to content
Main Site News Console

Sullivan & Cromwell Brings Specialized Legal Expertise to Contract Review

· OpenAI Translated
OpenAILLM

October 15, 2026

In collaboration with OpenAI, Sullivan & Cromwell developed Agreement Analyzer, an AI-powered tool designed to identify potential issues in draft agreements.

  • Organization Size: Enterprise
  • Region: North America
  • Industry: Professional Services
  • Product: ChatGPT

For Sullivan & Cromwell lawyers, reviewing a draft agreement involves weighing terms against relevant precedents, applying firm-tested strategies, and keeping client priorities front and center. The firm’s Agreement Analyzer brings these elements together into a unified review workflow, generating insights and redline suggestions that lawyers can review and refine.

“I think AI will… allow lawyers to focus on the areas where their judgment is needed most… to really zero in on the issues that are most critical to the client.”

— John Savva, Partner, Sullivan & Cromwell

Bringing Firm Expertise into Agreement Review

While AI was initially used primarily for repetitive, administrative tasks like drafting email replies or managing calendars, Savva noted that recent leaps in AI analytical capabilities have fundamentally shifted how the technology is applied. “Now, we see far more opportunities to apply it to challenging legal work, such as in-depth analysis of complex agreements,” he said.

Agreement Analyzer helps lawyers compare corresponding terms across different precedents and trace how those provisions interact with one another.

Sullivan & Cromwell partner John Savva initially built a single-user proof of concept (PoC) for Agreement Analyzer using Codex. He then collaborated with OpenAI to improve output quality, enhance efficiency through parallel processing, and scale the application into a multi-user tool for the firm’s attorneys.

Agreement Analyzer deeply integrates Sullivan & Cromwell’s workflows, curated firm precedents, negotiation playbooks, and ChatGPT Enterprise. After uploading a draft agreement, lawyers can supply selected precedents, playbooks reflecting firm and client positions, and market practice context. The tool performs multi-pass reviews from multiple perspectives, synthesizes the results, helps surface potential issues, and delivers a review memo alongside suggested redlines (tracked changes) for lawyers to review, accept, or reject.

“This is not about replacing lawyers; it’s about giving them a tool to conduct deeper analysis and surface more evidence and precedents, ultimately delivering a higher-quality work product for clients.”

— John Savva, Partner, Sullivan & Cromwell

Making Sources and Uncertainty Transparent

Savva emphasized that full transparency is essential for an AI tool’s output to be trusted. “If an AI tool makes an assertion about market practice, you have to be able to see the source of that information,” he noted.

Savva described an application architecture where primary agents draw conclusions while other agents challenge them through an adversarial process. In one prospective use case, specialized agents review an agreement in parallel—checking for deviations from market practice, risk allocation issues, inter-clause dynamics, and ambiguities that could pose litigation risks. The output clearly distinguishes between conclusions that withstood scrutiny from peer agents, rejected findings, and unverified issues flagged with caveats. Lawyers can then carefully evaluate these points, their supporting rationales, and the cited precedents before deciding which items should be reflected in the final redline.

What’s Next

Sullivan & Cromwell continues to explore custom tools tailored to specific legal workflows. In Savva’s view, stronger analytical capabilities open the door to supporting a wider range of research and drafting tasks.

Savva specifically pointed to the breakthrough in reasoning delivered by GPT-6 Astra: “Astra has really made a substantive jump in the quality of reasoning, which was previously a bottleneck,” he said.