Machine Learning in Legal Tech

Machine Learning
in Legal Tech

Claire Williams | November 20, 2018

Machine learning has become one of the most controversial and polarizing concepts, not only in the legal industry, but the broader enterprise software space. Years of premature and/or misguided marketing has groomed society with inaccurate expectations of machine learning software. With a ‘zero-room-for-error’ standard, this misrepresentation has caused the legal industry to become especially apprehensive of the ‘hype’ around new legal tech.

However with recent advancements in technology and more experienced technologists entering the game, companies are becoming more curious about machine learning software and its obvious and exponential benefits (workflow efficiencies, new revenue streams, bigger/better projects, etc).

Instead of summarizing opinions and boiling it down to a ‘marketing’ blog post, we sat down with three Heretiks, who spend their time passionately working to empower their users through their technology, to talk to each other about the state of machine learning in legal tech. Each has extensive experience in both the legal and tech industries. Jannie Chang (Data Scientist) and Rishi Khullar (Director of Product Management) are former lawyers turned technologists, while Charlie Connor (CEO & Co-founder) has spent his career in B2B enterprise software, most notably at Relativity.

We started by asking them the loaded, but basic question – why do you think machine learning such a polarizing and controversial concept, especially in the legal industry?

JANNIE CHANG

Data Scientist

I find that at their core, the legal industry and data science are just polar opposites. The legal industry generally makes money through billable hours. In contrast, the purpose of machine learning is to optimize everything, save time, make things faster. When you put the two together, they’re inevitably going to butt heads with money being tugged one way or the other.

RISHI KHULLAR

Director of Product

Do you feel like clinging to billable hours is like Blockbuster clinging to late fees?

JANNIE CHANG

Data Scientist

Yes.

RISHI KHULLAR

Director of Product

I don’t believe the law firm of the future will be doing this drudgery work and will be more focused on strategic work that they’re charging higher rates for and hiring less associates. Is that a law firm that is actually more profitable?

CHARLIE CONNOR

CEO & Co-founder

I don’t think that it is just a legal problem. In general, there is a lack of education on what machine learning is, so part of this concept polarization is that people are either already invested in machine learning that isn’t doing what they thought it was going to do, OR they think that it’s just a fad, which is interesting.

When I was at the World Economic Forum this year, the most common question was, “Is AI a fad or is it going away?” The reality is, it’s in everything you use… I think an additional problem for legal is the historical business nature. We’ve experienced this ‘easy button’ problem ourselves.

JANNIE CHANG

Data Scientist

I agree that a lot of the fear of AI and machine learning stems from the fact that most lawyers don’t know what it is. If they had a better idea of what it is, what it does, and how it can help them, then lawyers can find a solution for how to incorporate new tech into their existing business model.

Everyone agreed that the most pressing issue within the legal industry is the misguidance around what machine learning actually is and this is where the hesitation stems from.

CHARLIE CONNOR

CEO & Co-founder

Do you think it is a right-of-passage problem? Do you think people say, “When I was an associate, this is what I had to do?” And they just think ‘those damn millennials!’ Could that also be it?

JANNIE CHANG

Data Scientist

I remember that’s what it was like before the economy crashed, when there was plenty of work for all the lawyers out there. Now there are too many lawyers and not enough work. Billable hours are becoming a scarce resource that everyone is trying to scrape up. To be honest, technology has been shaving time off most lawyers’ billable hours for a really long time and it has only accelerated with machine learning. I think the legal industry essentially built its careers and finances around an antiquated idea that they refuse to let go of.

RISHI KHULLAR

Director of Product

The antiquated idea being that billable hours is the best business model for law firms?

JANNIE CHANG

Data Scientist

Yes.

CHARLIE CONNOR

CEO & Co-founder

I think it is a mix of things. They could use machine learning to remove a ton of inefficiencies, like tracking hours. But they could also have more billable hours, because they spend so much time just tracking their hours with how bad their e-billing is. They’re missing revenue on the table. So it seems like an interesting thing where there are so many more unknowns. If you could just say to them, “The way we are using machine learning would allow you to use this for your clients’ invoices, or projects or for your receipts and you would just cut and paste the pattern that they are usually doing these days.” Here’s a problem too – I think that everything is archaic.

Rishi and Jannie both believed that the ‘law firm of the future’ will be focused on efficiency and stream-lining workflow. However, this could be beneficial for the legal industry because it could potentially help identify new paths of revenue.

JANNIE CHANG

Data Scientist

I don’t think that it is the quality of the work that the lawyers are questioning, if you ask any lawyer, they’re always going to have things on their plate that they do not think is worth their time or energy, but it’s still on their plate. Essentially what’s happened is the market has become oversaturated with lawyers. There are too many lawyers out there and not enough quality billable hours. So then you have these lawyers billing for time on work that is beneath their skills and experiences. Sorting documents, photocopying, pulling files, simple tasks like that. The smaller the firm tends to have more mindless tasks, but all this stuff adds to revenue.

CHARLIE CONNOR

CEO & Co-founder

Yeah, and high margin revenue!

RISHI KHULLAR

Director of Product

But I think that there’s higher margin revenue with more strategic work. Also, it seems like corporations are getting more savvy to getting billed high rates for simpler type work.

CHARLIE CONNOR

CEO & Co-founder

That’s where I think the smart groups are realizing how much money they can collect by organizing their clients contracts. A lot of enterprises are getting more organized upfront, so there won’t be that much trouble.

JANNIE CHANG

Data Scientist

I don’t think law firms intend to dupe people. But with their most common and longest lasting business model, there’s no incentive to advance, get better, change…

RISHI KHULLAR

Director of Product

But that’s why I brought up Blockbuster because I think clinging onto the argument ‘let’s be more inefficient because we want more hours billed’ is not building a future. It’s not building a future law firm or moving client value with an obsession over customers. It’s basically like, “We’ve been duping people, let’s continue to do so and let’s not let other technology bar us from duping people.”

JANNIE CHANG

Data Scientist

It’s one history’s older professions.

CHARLIE CONNOR

CEO & Co-founder

People are trying to avoid risks. People always want to blame somebody else. There will always be that ‘should I sign this?’ moment. In the ‘High Growth Handbook’ by Elad Gilt, it says that one of the most important things to start hitting growth is to find a lawyer you can trust. The good GC’s are strategists.

JANNIE CHANG

Data Scientist

Oh I agree. Technology will never replace lawyers.

CHARLIE CONNOR

CEO & Co-founder

There’s another Economist article I read that said something about how everyone thought eDiscovery was going to kill so many law jobs, but it actually created more jobs because people were going to court more often. They were no longer saying, “The discovery is too much. let’s just settle.”

This is where I think the Big Four is changing, because they’ve started realizing law is not too different from how they audit finances.

Law firms are starting to wake up and say wait a minute, a financial auditing group is in our area?

People want to be risk adverse, it’s just on the law firms to see where there are more risks that they can council on. I don’t think it’s the billable hours part, I think that it’s realizing, “Oh my god, my whole life has been built to follow the rules, and now you’re telling me to be successful I have to break those? I would have to be like a Heretik, you know?” I just think that the reality isn’t around billable hours, I just think it is a deeper-rooted education problem, a lack of creativity, maybe they’re tired.

JANNIE CHANG

Data Scientist

I do agree to a certain extent. I think the legal industry needs to shave some fat to be honest. It’s over saturated with too many lawyers.

CHARLIE CONNOR

CEO & Co-founder

I would say something about what’s happening in the alternative legal service provider space is the small ones are being eaten up right now.

JANNIE CHANG

Data Scientist

For the larger ones though, I think that machine learning is going to be a huge asset. Once they understand it better, you’re going to find that the legal industry is still struggling to keep up with technology in terms of understanding it, explaining it, and using it. Once the fat has been shaved off, you’re going to find the remaining attorneys very anxious to get their hands on this, because essentially it cuts down all costs and the more cuts they make, the higher the profit margins.

RISHI KHULLAR

Director of Product

I think there is also a weird focus on machine learning. It’s not like machine learning is the only solution that is going to help in a contract review workflow, so there is all this focus on AI and machine learning. User experience, good design and solving the problems. People just jump to the solution of machine learning as being this all-encompassing thing, when we think about just as a tool in our tool set to solve some end problem for the customer. I think it’s just unfortunate that we are always focused too heavily on the machine learning side of it.

The conversation then transitioned into how startups like Heretik can prove their legitimacy in the industry, and how startups can help with confusion and misunderstanding around machine learning.

RISHI KHULLAR

Director of Product

I think for one, table stakes is to be honest and transparent about what you can and cannot do. When it comes to machine learning, there is a bit of overpromising in terms of the amount of data you need to train or how much time it will take to train a model, how many things it will extract automatically. Customers really appreciate that honesty, especially given the market. Also, you gain wider adoption by just stacking small wins with customers.

If you can say, “Hey we did this project with this new startup, machine learning was a part of it, and we ended up saving money or making more money and they solved our problem with this small case, we will use them again and again!” It’s not about wanting to be that check box for law firm or a corporation and having some cool new machine learning technology.

Jannie strongly believes data scientists are going to be crucial for development in the legal industry. Clients are going to ask why data scientists are no being used in law firms to fill these voids.

JANNIE CHANG

Data Scientist

I feel like the legitimacy comes from accessibility. Essentially machine learning has a lot of potential to solve a lot of problems that we haven’t even thought of in the legal industry. The problem is that there is very little overlap between the legal domain knowledge and the machine tools out there. To be honest, there is already a massive shortage of data scientists in the industries other than legal. Getting more data scientists into the legal space is probably going to be a big hurdle that won’t be tackled for a very long time. That is a gap that we can fill by making machine learning super accessible by giving these tools to people who are already starting to develop the tech experience, like eDiscovery professionals. I think that would quickly legitimize our tech in the legal industry.

Everyone also agreed that the most beneficial way for startups to be recognized and legitimized is by understanding that relationships within the legal sphere are extremely valuable.

CHARLIE CONNOR

CEO & Co-founder

For startups, this can be a way more tactical like both of you said. I think a startup has to come in with the mindset to do one thing very, very well and find the right partners that are going to start easing the technology in. I think that for many of Heretik’s first projects, if it was solely led by us, we would have never gotten through procurement or security. There was also the familiarity of a tool our clients already had (Relativity).

To your point, startups have to be very transparent with what they can and cannot do. They have to also be really solving the problems kind of methodically.

I always think about how everyone wanted us to just do data extraction right away and thinking back on it now how we unitize the contract based upon what we track with the data. And then really leveraging that with the right partners. There are companies we’ve talked to that have 35 year relationships with our clients. This reinforces that idea of trust with their council.

That’s why I think lawyers who lead legal tech startups start off well, because they have this validation or relationship with the legal industry already. In the longer game, as a technologist, we have to come in with people that are known and respected in that practice. It would be great for more data scientists to come in to the legal space, but what could be more intimidating for both sides?

RISHI KHULLAR

Director of Product

I think that contract unitization is a perfect example. If we had just focused on machine learning as a solution or AI as a solution, we might have gone the route of just auto-extracting as many terms as possible, even though there are companies out there doing that very well. Instead, we talked to customers and realized that one of the biggest problems is the grouping of amendments to master agreements and knowing final source of truth. From that problem, we’ve learned that machine learning can help us with the solution, but we didn’t approach it backwards, which I think makes it a good example.

CHARLIE CONNOR

CEO & Co-founder

Ultimately, I think our job is to help our clients win projects they never thought they could win. We have to do that by building their trust, being transparent, and spending time with them to educate them on what is possible and what is not. That to me is the most exciting opportunity. Watching our clients go to Adam more frequently now, because he has built that trust and they really understand the answers he is giving.

When you look at really successful tech companies, especially B2B, that trust has been established to go after something. Then clients can trust that we are not going to be the experts, but we ARE going to get them what they need to really exceed the expectations of their client.

That’s where the ability to slowly start adding value, on things like unitization or segmentation, that starts giving our clients the comfort to come to us and say ‘let’s tackle this thing I’ve wanted to for a really long time.’ That’s what I think our role is. I think instead of saying, ‘Here’s this thing that’s fully automated,’ startups should be saying ‘How do we work together to do something you didn’t think you could do?’ That’s what I believe our arc; tackling really big problems they didn’t think were possible. Like reviewing 80,000 emails in two days. People never thought that was possible and now it’s a joke if you can’t do it.

The fifteen minute conversation wrapped with the group reiterating Heretik’s vision of empowering professionals at their trade, and how empowerment is the most important word in that sentence. We are building an unbreakable trust with our clients by being transparent, holding each accountable, and sticking to our core values. While there still may be some stigma around machine learning in the legal industry, we at Heretik are committed to empowering, educating, and supporting our clients to make sure they are wildly successful and confident with our solution.