Real Estate Law Blog

New York Real Estate Journal published an article on November 25 authored by Co-Chair of Olshan’s Real Estate Law practice Thomas Kearns entitled “Legal AI for Real Estate Transactions” (or here for NYREJ subscribers).

In the 1980s we papered real estate transactions using custom drafted riders attached to forms printed by publishers like Blumberg, REBNY and others. As a young lawyer I grew frustrated with trying to keep track of precedent riders, agreements, and checklists, so I created a form bank for use by the department. In many ways our practice has changed significantly. In other ways, not so much. Despite modern document management systems, it is not always easy to find helpful precedent documents. But modern AI tools are able to search large repositories of documents like our set of precedents combing for answers to questions. Modestly priced consumer level AI tools do the same task with the massive publicly available database.

I have been on our firm’s AI committee, and we have been screening numerous AI products to help in our practice. The hope is that these products will help us mine our vast database of transactions by scanning millions of pages of documents in seconds and answer questions or even preparing drafts of suggested clauses based on our long history of negotiated agreements. Two recent experiences are illustrative of the issues and benefits of these products.

Our summer associates ran a test on publicly available AI engines such as ChatGPT, Gemini, Grok and others. The test was a request to draft a paragraph under New York law to insert into a limited liability company agreement regarding indemnification of an LLC manager. I purposely picked a hard test since New York indemnification law can be complex and confusing but it’s also a topic that is important to people who serve as managers of LLCs. The AI engines all failed in the task. None picked up the nuances of New York indemnification law particularly as it came to legal fee reimbursement and legal fees to enforce the indemnity. While I would not use the word “hallucinations” to describe the results, the results were weak.

In perhaps a fairer test, over the years I had developed a checklist of issues in LLC agreements drafted by sponsors of a commercial real estate deal where we are representing a minority, non-controlling investor in the deal. When a colleague had a similar deal, he was able to use an AI tool we are testing to ask how the draft addressed the issues in my checklist and the answer was very helpful. Most AI tools for lawyers also allow for the law firm to create a subset of documents that the AI tool should look at for answers. Some tools call them golden standards, others call it a folder, a vault, a playbook or similar names but the concept is the same – a curated database of trusted documents. After all, our extensive database of documents also includes poorly written agreements by other lawyers that we have because they were litigated and documents drafted before new court decisions were issued or new statutes passed which changed the law.

The form databank that I started 40 years ago may now come in handy as it may be a logical place for us to start to create those golden standards. So while much has changed over the decades, the need for us as a group of commercial real estate lawyers to collaborate on legal issues, drafting standards and market trends continues. In fact, the AI tools we select will only enhance our ability to mine our significant database to help serve our client’s more efficiently.

But going back to the test we ran re New York indemnification clauses, real estate professionals should be careful about using publicly available AI tools to answer complex legal questions. These tools can certainly be a help in some circumstances but so far, they are nowhere near being able to replace experienced lawyers on complex issues.

Thomas Kearns is a partner in the real estate department with Olshan Frome Wolosky LLP, New York, N.Y.