Recently, we were joined by staff for a casual chat about all things AI, machine learning, and contracts, hosted by Kelly Williams, Senior front-end developer, and joined by Dr. Kat Hempstalk, VP of Data & Analytics, and Ali Meredith, Analyst in Data & Analytics.
What’s different about LawVu’s approach to contract management and machine learning from what you’ve seen in the past?
Alli: I think LawVu’s main difference is that it’s a very holistic platform. And it’s designed for in-house. There are a lot of tools out there for legal tech that are designed to solve just a singular pain-point, and they’re designed for law firms rather than in-house counsel. They’re designed to solve a particular problem that a law firm will have, like building contracts.
Whereas with LawVu, because we’re holistic, we solve a whole heap of problems. It means we can take what we’ve done with matters or with billing or with anything, and we can apply that into contracts. It also means that when we’re trying to find ways to manage contracts better, we’re using what we’ve learned from managing matters and we’re pulling that knowledge into how we manage contracts. I think that’s the key differentiator for LawVu.
What’s the most exciting thing happening in this space right now?
Alli: The thing that really gets me going is contracts. There’s nothing else for me – I love it. There’s so much we can do. A lot of what lawyers do is just literally reading contracts, policies, legislation regulations. They spend a lot of time deep in documents when they could be doing higher-value work.
And that’s perfect for machine learning and AI because we can make it easier, we can take the information across a set of documents and say, these are the key documents you need to read. These are the key paragraphs. You don’t need to bother, or you don’t need to read so closely all the rest.
I think the most exciting thing for me is applying that across such a broad range of documents, which will help lawyers spend less time doing very boring things.
Dr. Kat: Something I saw a few years ago now was the concept of style transfer.
This means being able to take an image that’s made by someone famous, and apply their style to a photograph that someone like you or I have taken. And so suddenly, you can take a photo with an app and it looks like a painting that’s authored by van Gogh, rather than a photo we’ve taken with phones.
I think that there’s a relevant application for the law and documents, where you can have a generic form document where it’s standardized fields that are clauses from a legal standpoint and apply the style of your writing or mine to that document, so that it sounds like you or I wrote it, without changing any of the actual meaning behind anything that’s there. I don’t believe anyone’s done it yet, but it sounds pretty exciting for me and I’d like to give that a go.
That would be pretty cool.
What are the biggest barriers for AI and machine learning in the legal tech space?
Dr. Kat: Trust is the biggest one. I’ve worked in healthcare and that’s often life or death – a machine is making a decision and that doctor has to trust what the machine is saying. And so it’s making the AI explainable and building that trust that it is actually doing the right thing. You make the decisions that the machine is suggesting. And, it usually takes time – trust isn’t a given from the beginning.
Ali: There are so many challenges, which one? I think one thing, one thing that strikes me as a huge challenge and it’s not about the technology itself is adoption. And the legal system, in particular, is really ill-suited to adopting new things because lawyers are very risk-averse.
They don’t like change. Change is scary and changing something could lead to risk. If you put this AI system in place and it misses something and that ends up going to court, the courts have never seen that before. You don’t know what’s going to happen, whereas if you do it the old way, even if it costs you double the amount it’s safe, even if you need 40 people in a room to do it instead of two, at the end of the day maybe it’s on the safe side.
So, a huge thing we have is like Kat said, building trust and convincing people that our tool does work.
How do you do that? How do you get the buy-in?
Dr. Kat: Years ago, I worked on a camera system for cows and one of the farmers had concerns that the camera was doing the wrong thing. He saw a cow that he didn’t think should go back out to the paddock heading back out. They chased the cow down with his four-wheeler, herded her up and it sounded like it wasn’t a great experience for him to have to do that. But, when he got the cow back up to the shed and actually looked it over, it turned out that the camera system had been right, and he was wrong about that.
Up until that point, he was suspicious of it and wanted to double-check the results. I think the same thing will be (understandably) true of lawyers, that same risk-aversion will come through on some of those things.
Once they see it working for them and there’s a decision that could go either way and actually, the machine gets it right straight away, that will help build that trust and adoption. Once you start to trust it, you start to tell your friends. Word of mouth is powerful. It brings people along that journey.
In other industries, there’s a lot of innovation happening already. Do you see the same being possible with the legal industry?
Dr. Kat: Language is hard. And a lot of the legal space is to do with language, whether it’s suggesting it, reading it, searching it, or forming new words and meaning behind the documents.
And it’s one of the hardest things. Machine learning images are far more structured. If you ask anyone who’s learned to speak English from another language, they will tell you all the nuances to English. And we’ve got all the challenges of all those languages as well as interpreting the meaning behind the text. So there are huge opportunities there.
Are there any hints you can share with us about what you’re working on at the moment?
Dr. Kat: We can probably give a few hints. We’ve been working with a few of our customers who have kindly donated some of their data to help us get started on contract automation building. We’re looking at making it easier to use LawVu with respect to those contracts. We’re looking at what we can automatically determine from them and load into LawVu that will help them. Have anything to add, Ali?
Ali: We’re bringing magic. It’s very hard in the background, but hopefully, when it’s integrated into the LawVu platform it will help our customers immensely.