Sullivan & Cromwell partnered with OpenAI to create Agreement Analyzer, an artificial intelligence tool designed to accelerate and deepen review of draft agreements. The application combines the firm’s curated precedents, negotiating playbooks and ChatGPT Enterprise to produce analysis, a memorandum and proposed tracked changes that lawyers can inspect and modify.
How the tool is used
Traditionally, Sullivan & Cromwell lawyers review a draft agreement by comparing its terms to relevant precedents, applying the firm’s experience-based approaches, and keeping the client’s priorities in mind. Agreement Analyzer brings those inputs into a single workflow: after a lawyer uploads a draft, they can add selected precedents and a playbook containing firm and client positions and market-practice information. The tool runs multiple review passes from different perspectives, synthesizes outputs, identifies potential issues, and returns a memo and suggested tracked edits for a lawyer to accept, adapt, or reject.
The application also helps lawyers compare corresponding provisions across precedents and track interactions among clauses.
Development and technical details
Partner John Savva developed Agreement Analyzer initially as a single-user proof of concept built with Codex. He then worked with OpenAI to improve output quality, increase efficiency via parallel processing, and convert the prototype into an application usable by multiple lawyers concurrently.
Agreement Analyzer can employ specialized agents that examine a contract in parallel — for example, to detect market deviations, assess risk allocation, analyze clause interactions, or flag ambiguities that could affect litigation. The system distinguishes findings that remain supported after adversarial challenges by other agents, findings that are dismissed, and issues retained with qualifications pending verification. Lawyers can review the flagged issues, the supporting explanations and the relevant precedents, then decide which findings should lead to changes in the final markup.
Transparency and trust
Savva stresses that transparency underpins trust in AI outputs: “If an AI tool makes a claim about market practice, you need to be able to see the source of that information,” he said. The tool therefore exposes supporting sources and relevant precedents so lawyers can evaluate the basis for its conclusions.
Changes in AI use within the firm
Savva notes that AI was initially applied mainly to repetitive administrative tasks such as drafting email responses or checking calendars. Recent improvements in models’ analytical abilities, however, have broadened potential use cases into more demanding legal tasks, including deep analysis of complex agreements. He specifically points to advances in reasoning with GPT‑6 Astra: “Astra’s really moved things forward in terms of the quality of the reasoning that’s available, and that has been the bottleneck in the past.”
Next steps
Sullivan & Cromwell continues to explore tools tailored to particular legal tasks. For Savva, stronger model reasoning presents an opportunity to support a greater portion of the research and writing process — while the ultimate review, judgment and responsibility remain with the lawyers.
