Sullivan & Cromwell's Agreement Analyzer Brings AI Into Contract Review With Codex and ChatGPT Enterprise
Sullivan & Cromwell has built an internal contract-review tool called Agreement Analyzer that combines the firm's curated precedents, negotiating playbooks and ChatGPT Enterprise to produce a memorandum and proposed tracked changes for lawyers to review. The tool matters because it shows a frontier AI lab and a major law firm positioning AI on analytical legal work — comparing provisions across precedents and tracking how clauses interact — rather than on the administrative tasks that dominated early legal AI adoption.
What Agreement Analyzer does
Agreement Analyzer reviews a draft agreement against the firm's own precedents and against client-specific negotiating positions, then returns analysis and proposed changes that a lawyer can inspect and refine. The tool compares corresponding provisions across precedents and tracks how those provisions interact with one another.
The workflow is input-driven: a lawyer loads a draft agreement into the tool, then adds selected precedents and a playbook containing firm and client positions as well as information on market practice. From there, the tool performs multiple review cycles from different perspectives, synthesizes the outputs, helps identify potential issues, and returns a memorandum and proposed tracked changes. The lawyer can review, adapt or reject those outputs.
How the tool was built
Sullivan & Cromwell partner John Savva developed Agreement Analyzer as a single-user proof of concept built with Codex. He then collaborated with OpenAI to improve the quality of outputs, enhance efficiency through parallel processing, and transform the application into a tool that could be used by multiple lawyers at the same time.
The production tool integrates Sullivan & Cromwell's workflows, curated firm precedents, negotiating playbooks and ChatGPT Enterprise.
Why the firm says AI use is shifting toward analytical work
Savva describes a change in where AI is applied in legal practice. Early use centered on repetitive, administrative tasks such as drafting email responses or checking calendars. Savva attributes the shift to recent improvements in AI models' analytical capabilities.
"Now, we're seeing many more opportunities for use in challenging legal tasks such as deep analysis of complex agreements."
—John Savva, Partner, Sullivan & Cromwell
Savva frames the tool as an aid to legal judgment rather than a substitute for it.
"I think AI will . . . allow lawyers to focus on the areas where their judgment is most important . . . so they can concentrate on the questions that are most critical for their clients."
"It's not about replacing lawyers. It's about providing a tool for lawyers to do deeper analyses and find more authority and precedents to provide an even higher quality product for their clients."
Transparency and adversarial verification
Savva argues that the trustworthiness of an AI tool's output depends on full transparency about where its claims come from.
"If an AI tool makes a claim about market practice, you need to be able to see the source of that information."
—John Savva, Partner, Sullivan & Cromwell
Savva describes an application architecture in which agents develop conclusions and other agents challenge them in an adversarial process. In a potential use case, specialized agents review the agreement in parallel for market deviations and risk allocation, interactions between clauses, and ambiguities that could affect litigation.
The results separate findings into three categories: findings supported after challenge by other agents, dismissed findings, and unverified issues retained with qualifications. Lawyers can inspect the issues, their supporting explanations and the relevant precedents, then decide which findings warrant changes in the final markup.
What's next
Sullivan & Cromwell says it is continuing to explore tools shaped around specific legal tasks. Savva sees stronger analytical capabilities as an opportunity to support more of the legal research and writing process.
Savva points to advances in reasoning capabilities with GPT‑6 Astra as a driver of that shift.
"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."
—John Savva, Partner, Sullivan & Cromwell