16-17 June 2027 – London, InterContinental O2 | Magazine

LegalTech Diaries Volume 17

Raffi Isanians

CEO
Mage Legal

LegalTech Diaries Volume 17

Raffi Isanians

CEO
Mage Legal

You worked as an M&A and startup lawyer at firms like Kirkland & Ellis, Gunderson Dettmer and Goodwin, before founding Mage. What was the moment or series of moments that convinced you that diligence needed to be rebuilt from scratch with AI?

There was one moment. I was reviewing an agreement in a merger, and I asked ChatGPT to draft me a terms of service. This was back when it was GPT-3.5. I immediately realised it was 90% of the way there. That was the moment I knew the legal industry was going to change, and that it was the perfect time to leave.

The background is that I was an engineer before I was a lawyer. I had always been programming to make my life easier as a lawyer, even if it was just automating things I was doing on a diligence memo. So when I saw what that model could do, I wasn’t guessing at what was possible. I had been building toward it the whole time.

Going through Y Combinator is a different path from most legal tech launches, which tend to come out of established firms or bigger teams. What did that experience teach you that your years in Big Law never could?

Going through YC was the best thing I could have humanly done. There is a thing called lawyer brain, which is the way we are trained to think as attorneys. I still think that is immensely valuable from a product creation standpoint. But the reality is that being a lawyer does not teach you how to run a startup. They have very different ways of working. So for all intents and purposes, going through YC really felt like starting at ground zero, rewiring my brain on how to build a successful company while still maintaining the lawyer side of me as we built our product. They are two very different sides of the brain, and very few people can operate in the middle of that Venn diagram.

Diligence has always been one of the most labour-intensive, highest-stakes parts of a deal and exactly the kind of work everyone assumes AI should be able to handle. It’s also exactly the kind of work where mistakes are costly. Where do you think the industry is over-claiming on AI diligence right now?

Accuracy at scale. I think it is insanely difficult to have accuracy at this context level. A data room with 2,000 documents, or even 200, is extremely difficult to parse. It is a needle in a haystack problem, and anyone can look accurate on ten documents.

The one thing our product does exceptionally well, and why we excel over our competition, is our ability to search with a high degree of accuracy in insanely large data rooms. It is at the point now where the product is more accurate than the attorneys were on the deal. When we run it against past deals that have already closed, we find issues the attorneys missed. That is the test we invite firms to run against us.

Join 5,000+ legal leaders shaping the future of law for two days of insight, innovation and unmissable networking — now with your biggest saving!