Ropes & Gray details AI tool for deal reviews, with lawyers checking the work

Ropes & Gray is putting AI to work on the document-heavy first stage of acquisition reviews, while keeping lawyers responsible for the final advice. In an October 29 account published by OpenAI, the firm described a proprietary tool developed with the company to organize deal documents and draft diligence reports for lawyers to check.

The disclosure describes a development effort inside the law firm, not a public product launch. Its initial focus is material commercial contracts: the agreements lawyers examine while assessing an acquisition. Before a purchase closes, the buyer must investigate the seller’s financial and legal claims, looking for risks and liabilities. That can mean intensive document research under a tight bidding deadline.

Reading relationships, not just individual contracts

The tool starts by inventorying and categorizing a data room—the collection of documents supplied for a deal review. It combs through thousands of pages, identifies relevant documents and groups related materials by themes and issues. Ropes & Gray says this lets the workflow examine full relationships rather than treat every document as a separate item.

That organization is itself subject to inspection. Lawyers can check the groupings and categories to confirm that nothing has been missed. The workflow then produces summaries and a diligence report in the firm’s own format. The intended output is a structured first draft of the work, rather than only a set of answers to individual contract questions.

Ropes & Gray built a consistent query and framework around the steps its lawyers would normally follow. Those steps run from the initial review through summaries and higher-level review. The firm describes the resulting first pass as comparable to work by a capable junior-to-mid-level associate. That is the firm’s assessment, not an independent evaluation of the tool.

Our clients don’t pay us to review contracts. They pay us to make judgments based on what’s in the contracts.

Howard Glazer, partner and head of Practice AI at Ropes & Gray

Past deals provide the testing material

During working sessions, the firm’s lawyers compared test outputs with human-authored diligence reports from completed transactions. They used what the drafts got right—and where they fell short—to refine the assessment criteria and report format. This testing ties development to examples of finished legal work, rather than relying only on whether an AI-generated draft reads convincingly.

The account does not publish an accuracy score or a measured time saving for the diligence tool. It describes a refinement process, but does not quantify how closely its drafts matched the lawyers’ reports. That leaves the associate-level comparison as a qualitative description, without a published result readers can use to judge the size of the improvement.

A faster draft still needs a lawyer’s sign-off

Review is built into the workflow, not reserved for the finished report. Lawyers can inspect interim summaries and the underlying documents before anything reaches a client. Josh Lichtenstein, a partner and head of the firm’s ERISA Fiduciary Practice, says lawyers still need to review everything because they must stand behind the advice they deliver.

The firm’s stated rationale is to shift lawyers’ effort toward reviewing, refining and exercising judgment after the initial document work is done. Its longer-term ambition is to produce a detailed, issue-level diligence report in hours rather than weeks. OpenAI presents that timetable as an ambition, not a demonstrated turnaround time for the system.