Sullivan & Cromwell Develops Agreement Analyzer with OpenAI
Sullivan & Cromwell has partnered with OpenAI to create Agreement Analyzer, an advanced AI application designed to streamline the review of complex legal agreements by integrating firm precedents, negotiating playbooks, and enterprise AI capabilities.

Integrating Legal Expertise Into Agreement Review
When lawyers at Sullivan & Cromwell review a draft agreement, the process involves carefully weighing contract terms against relevant precedents, applying approaches rooted in the firm's accumulated experience, and remaining mindful of individual client priorities. To streamline this complex workflow, the firm partnered with OpenAI to develop Agreement Analyzer, an AI tool built to identify issues in draft agreements. According to information shared via [OpenAI News](https://openai.com/index/sullivan-cromwell/), the application brings these critical inputs into a single review workflow, producing comprehensive analysis and proposed changes that lawyers can inspect and refine. Further details are available from OpenAI News in the original source material.

Evolution From Proof of Concept to Enterprise Tool
Sullivan & Cromwell partner John Savva originally developed Agreement Analyzer as a single-user proof of concept built with Codex. Following the initial concept phase, he collaborated with OpenAI to enhance the quality of outputs, improve efficiency through parallel processing, and scale the application into a robust tool capable of being utilized by multiple lawyers simultaneously. While artificial intelligence was historically deployed primarily for repetitive, administrative tasks such as drafting email responses or checking calendars, Savva notes that recent improvements in AI models' analytical capabilities have fundamentally altered how the technology is utilized across the firm.
Noting the shifting landscape of legal technology, Savva points out that the industry is witnessing significantly more opportunities for AI application in challenging legal tasks, such as the deep analysis of complex agreements. Agreement Analyzer directly addresses this need by helping lawyers compare corresponding provisions across precedents and track how those provisions interact within a given contract.
Workflow and Multi-Perspective Review Cycles
The Agreement Analyzer seamlessly integrates Sullivan & Cromwell's proprietary workflows, curated firm precedents, negotiating playbooks, and ChatGPT Enterprise. When utilizing the tool, a lawyer inputs a draft agreement and supplements it with selected precedents, a playbook outlining firm and client positions, and relevant market practice information. The application then executes multiple review cycles from distinct perspectives, synthesizes the outputs, helps surface potential issues, and returns a formal memorandum alongside proposed tracked changes for professional review, adaptation, or rejection.
Transparency remains a central pillar of the tool's trustworthiness. Savva emphasizes that if an AI tool makes a specific claim regarding market practice, users must be able to verify the source of that information directly.

Adversarial Agent Processing and Reasoning Capabilities
To ensure thorough evaluation, Agreement Analyzer employs an adversarial process where specialized agents develop conclusions while other agents challenge them. In practical use cases, these parallel specialized agents review the agreement for critical factors such as market deviations, risk allocation, clause interactions, and ambiguities that could potentially impact future litigation.
The resulting findings clearly distinguish between points supported after challenge by other agents, dismissed findings, and unverified issues retained with specific qualifications. Lawyers retain full oversight to inspect the issues, review their supporting explanations alongside relevant precedents, and ultimately decide which findings warrant modifications in the final document markup.

Future Exploration in Legal Technology
Sullivan & Cromwell continues to actively explore innovative tools shaped around specific legal tasks. For Savva, these stronger analytical capabilities present a valuable opportunity to support broader aspects of the core research and writing process.
Highlighting technological advancements, Savva points to improvements in reasoning capabilities with newer models like GPT‑6 Astra. He notes that Astra has significantly advanced the quality of reasoning available to legal professionals, addressing a critical bottleneck that has historically limited AI integration in complex legal workflows.
Sources
- OpenAI NewsSullivan & Cromwell brings legal expertise into agreement review
Continue chronologically





