Seattle-headquartered contract lifecycle management vendor Lexion today (20 December) unveils its AI Contract Assist tool, which uses GPT-3 to help lawyers with contract review and drafting.
Contract Assist is a new addition to Lexion’s Word plugin that helps lawyers draft, negotiate and summarise contract terms. It can automatically generate a first pass clause, which it says can be refined as needed. The new offering suggests clauses from the internet or a bank of prior contracts; produces redlines from a description of the new stance needed; and summarises clause language.
Speaking to Legal IT Insider, Lexion’s CEO Gaurav Oberoi said: “Imagine you’re working on a typical commercial contract between two companies. Maybe you are selling your product to them, or maybe you’re buying a product. It’s a vendor and maybe you’re looking at this vendor contract and you say, ‘there’s no mention in here about logo rights and usage. But our company is sensitive about that and I want to put a clause in here that says that they don’t have the rights to use our name or logo.’
“Lawyers don’t typically draught this kind of thing from scratch. They go and look at examples that they’ve either done before in their contracts, or they’ll go to the Internet and say, ‘hey, find me a clause for logo use.’ They’ll copy and paste but now they’ll be able to just type in ‘give me a clause for not using my logo and name,’ press enter, and we will then propose a clause right there within Word.”
GPT-3 is the latest language model from OpenAi that has been trained on the entire corpus of the internet, including millions of contracts. It leverages deep learning to generate human-like text, but that text is not always right, so Legal IT Insider asked Oberoi how users can be confident in the accuracy of the text they are using. “What we’re really doing is providing a trained lawyer with an assistant, we’re not providing a trained lawyer with ‘hey you don’t need to look at this, it’s automated,'” he said. “We’re saving them a bunch of time by saying, ‘here is an initial starting point.’ It may be the equivalent of what a junior attorney on your team may write, including the possibility of being a little bit off, but it saves you starting from scratch or searching the internet.”
The idea of encouraging lawyers to turn to the Wild West of the internet for contract drafting might be enough to send some KM heads to their graves. There are also still many questions around the likes of how copyright laws will keep up with GPT-3 and how to properly regulate it, but Oberoi says: “People are already using the outputs of these model and we think there is an opportunity to do the same with contract clauses, it makes a tonne of sense.”
While Contract Assist is not the end-to-end contract negotiator that we have seen lately in the likes of DoNotPay, it lets you select a clause to rewrite or update by proposing redlines. Oberoi says: “Again our approach is you, the human lawyer, with your training and expertise, are still ultimately shipping this contract back to the counterparty to review. What you are really getting is an assistant that is saving you a bunch of time.”
He adds: “It’s hugely time saving and it seems magical when you use it but it can go off the rails sometimes, so it’s not something you’re going to not review and say ‘let’s just go for it.’ What you’re saying is ‘I’m here watching you.’ Maybe a good analogy is driverless cars. We’re not yet at the stage where you can take a nap.”
Lexion has had Contract Assist in beta with existing customers and Oberoi says: “The feedback we’re getting is ‘wow, this is incredibly helpful. It’s saving me a bunch of time and it’s really easy to use of the shelf, plus it’s right there in Word, so it’s not like I have to go to another web page.”
As Lexion gains more customers using the tool, Oberoi says that the company will be able – with permission from customers – to analyse how it is working. “The fact that we’re in an end-to-end contract management platform means that we are capturing all this data and that gives us more fodder to feed into these algorithms that we will be able to monitor over time,”he said.
caroline@legaltechnology.com
Interesting blog, keep it up!