> ## Content Index > Fetch the complete content index at: https://www.notatechguy.com/llms.txt > Use this file to discover other available public pages before exploring further. # Ropes & Gray builds AI diligence tool with OpenAI - URL: https://www.notatechguy.com/ropes-gray-builds-ai-diligence-tool-with-openai/ - Published: 2026-10-29T20:35:52.000Z - Updated: 2026-10-29T20:35:52.000Z - Description: Ropes & Gray is building an AI tool for transactional diligence with OpenAI, but lawyers still review every output before advising clients. - Author: Marcello Babbili - Tags: Technology & AI, OpenAI, Google Ropes & Gray is building an AI tool for transactional diligence in partnership with OpenAI, according to a 29 October announcement on OpenAI's website [S¹](https://openai.com/index/ropes-gray-ai-diligence?ref=notatechguy.com). Lawyers at the firm still review every grouped document, summary, and report before advising clients [S¹](https://openai.com/index/ropes-gray-ai-diligence?ref=notatechguy.com). The announcement is the sole source for the partnership. No independent reporting has confirmed the tool's maturity, the model behind it, or any measured result. **My read:** This is the thinnest evidence pack I have worked from in a while. A single promotional post from one of the two parties, with no numbers, no model name, and no third-party corroboration. What I find credible is the workflow design: lawyers reviewing AI output before client advice is exactly how a regulated profession would adopt this. What I do not buy yet is any implication about efficiency or cost savings, because none are offered. The real signal is OpenAI's continued push into vertical professional tools, not the specifics of this one deal. Transactional diligence is the document-heavy phase of mergers and acquisitions, where lawyers comb through thousands of contracts, employment agreements, and regulatory filings under deal deadlines to flag risks a buyer should know about. It is repetitive, high-stakes, and billed by the hour. A natural target for automation, but also one where a missed liability clause can cost millions. The tool organizes related documents, drafts summaries, and assembles reports that attorneys then examine before anything reaches a client [S¹](https://openai.com/index/ropes-gray-ai-diligence?ref=notatechguy.com). That workflow mirrors how AI is entering other professional services: the machine does the first pass, the human does the judgement. AI handles volume, humans handle decisions. ### The announcement names no model and reports no benchmarks No specific OpenAI model is named. The firm could be using a frontier model, the open-weight gpt-oss models OpenAI released in mid-2025 [P⁴](https://github.com/openai/GPT-oss?ref=notatechguy.com), or a custom-built system. The announcement does not say. There are no benchmark scores, no error rates, no before-and-after time comparisons. Whether the tool has been deployed on live deals or is still being tested internally is not disclosed. These claims rest solely on the vendor's own announcement; no error bars, no released code, and no independent benchmark have been published. For a law firm, that silence is partly structural. Client confidentiality rules make it difficult to publish details about tools that process deal documents. But it also means the public record contains no way to verify whether the tool works as described. ### Three questions any legal-tech team should ask before copying this approach A law firm or legal-technology team watching this space should ask three questions before drawing lessons from the Ropes & Gray announcement. What model is doing the document analysis? How does the firm handle hallucinated summaries, where the AI invents clauses that do not exist in the source contract? And what is the lawyer's review workflow: does every AI output get a line-by-line check, or is the review sampled? The announcement, as published on OpenAI's platform [S¹](https://openai.com/index/ropes-gray-ai-diligence?ref=notatechguy.com), answers none of these. It is a direction signal, not a proof point. The gap between announcing an AI partnership and shipping a tool that holds up under professional scrutiny is where most of these efforts stall. For a junior associate on a due-diligence team, the practical change would be receiving AI-grouped document clusters and machine-drafted summaries to review instead of manually sorting through raw contract batches — but only if the firm adopts a tool like this, and the announcement does not confirm deployment. The next checkable milestone would be a case study, a conference presentation, or a regulatory filing from Ropes & Gray describing the tool in live operation. Until then, the partnership is a confirmed plan with no confirmed results. --- *Sources: [S1 — How Ropes & Gray is developing AI for transactional diligence](https://openai.com/index/ropes-gray-ai-diligence?ref=notatechguy.com) · [S2 — How Ropes & Gray is developing AI for transactional diligence - OpenAI](https://news.google.com/rss/articles/CBMiXkFVX3lxTE5sdFBJTFB3bDUwOEhVSnFNZ3BvS2tkTVFkY2dHNzk1QWNJMU1oSTVFSjRpR0g4akZsNDdjcmZmVDR0M2Q1QmNwcERKZ2NYdXE2NVNXaU5MaURYaXBHTGc?oc=5&ref=notatechguy.com) · [P3 — How Ropes & Gray is developing AI for transactional diligence | OpenAI](https://openai.com/index/ropes-gray-ai-diligence/?ref=notatechguy.com) · [P4 — openai/gpt-oss](https://github.com/openai/GPT-oss?ref=notatechguy.com) · [P5 — mjbommar/gpt-takes-the-bar-exam](https://github.com/mjbommar/GPT-Takes-the-Bar-Exam?ref=notatechguy.com) · [P6 — graphdeco-inria/gray](https://github.com/graphdeco-inria/gray?ref=notatechguy.com)* --- *Written from 6 sourced items, 5 of them primary.* ## More from Not A Tech Guy - [OpenAI reveals layered monitoring to catch rogue AI agents](https://www.notatechguy.com/openai-reveals-layered-monitoring-to-catch-rogue-ai-agents/) - [Gemini app now edits PDFs and docs inside chat](https://www.notatechguy.com/gemini-app-now-edits-pdfs-and-docs-inside-chat/) - [IBM Granite Switch 4.1.1 patches adapters, not base model](https://www.notatechguy.com/ibm-granite-switch-4-1-1-patches-adapters-not-base-model/)