Intro:
Welcome to the Contract Lens Podcast brought to you by Malbek. In this podcast, we have conversations with contract management thought leaders and practitioners, about everything contracts and its ecosystem. Today's episode focuses on one of today's hot topics, AI based contract management, discussing exactly what AI can do for contracting. The discussion is led by Malbek CEO, Hemanth Puttaswamy, and he is joined by Teju Deshpande, founder and CEO of Oya Solutions, a boutique consulting firm, focused exclusively on delivering innovative contract management solutions. Teju has spent over 25 years in the technology and legal services industry, helping leaders improve effectiveness and efficiency through process technology and data optimization. So now it's time to relax, grab a glass of wine, and let's talk contracts.
Hemanth:
Hi, Teju. Today, we're going to talk about AI in legal tech. With me, I have a wonderful guest, Teju Deshpande, who is the founder and CEO of Oya Solutions, a boutique firm focused on contract lifecycle management solutions. So, Teju, welcome, how are you today?
Teju:
Hi, Hemanth, I am thrilled to be here, and excited to kick off this discussion.
Hemanth:
Awesome, Thank you, Teju. So, Teju, just to start up this discussion, which is AI and ML in the legal tech, let's start with one interesting question, that comes in pretty much every industry. Are businesses ready and open to adopt AI in the legal space?
Teju:
That's an interesting question, because people think about AI differently. If you think about AI solutions, businesses particularly legal departments, have been using AI for almost a decade in various shapes or forms. The first version of that really came out around e-discovery, for Legal departments in particular, where there's large volumes of documents to be reviewed. That technology, in a concept, translated to law firms doing extractions of metadata and contract summaries for due diligence.
Teju:
Companies like KIRA and eBrevia really started out with concepts of, "Can I extract data for due diligence purposes?" Which then led to some of the innovations that we see today, and there are many, many companies that now offer what you would call standalone extraction tools, that link with various different systems. And they have other innovations, if you will, in terms of AI that relates to negotiation and review. There are companies that are platforms, like Malbek, that actually embed AI, because you want to make sure that it's part and parcel of the entire solution, not just a standalone extraction tool on the side.
Teju:
So because there are different variations, businesses have been ready, I think, for the extraction element. They're still not quite ready in terms of the negotiation and review component, because their playbooks are not defined yet, is what I think, in the contract management world. Has that been your experience as well, Hemanth?
Hemanth:
Very similar, Teju. Yeah, you're absolutely right. The extraction, analytics, discovery, these are some of the use cases that's been there quite some time, as you said. The AI auto review, there are bits and pieces that businesses and customers are open to, and that's something that we are working on, and we're very excited on those areas. I believe in the next year or so, in my opinion, they may start adopting that. What do you think? Because you’re on the review side, do you see that as a trend?
Teju:
I think so. And if I think about the contracting world in particular, people start out with, "I want a simple repository first. Then I want to enable some level of process automation and templates." Then you get to the, "I want advanced analytics." And then comes AI as the roof on the top of the house. What I have seen as fundamental shift, and this is what I talked to a lot of my customers, is to follow that same contract, but instead of a simple repository, replace it with a smart repository, which has AI as the foundation.
Hemanth:
Yep.
Teju:
And so, what you want to look for are platforms that are thinking, "AI first." Then you don't necessarily have to extract and have your AI read 150 terms. You can start by extracting five or 10 key operational terms, but you start there. And so anything that you apply from a process automation standpoint, applies universally to the purpose of existing documents, as well as go-forward documents. So I think that's the shift that I'm seeing, which will then enable, like you rightly said, the review component. So if you have a basis for automated extraction and analysis, then you can build an automated playbook. And then you can build on that for automated review.
Hemanth:
There is huge discussion, "AI is going to take away jobs." And there are there are people who are thinking that AI is going to replace the lawyers or the legal team. What's your perspective, Teju, from what you're seeing?
Teju:
That’s an interesting fear, for lack of a better word that pops up every time there is automation, or artificial intelligence related conversations. We heard it in the e-discovery world, we heard it when contract extraction and summarization came along initially. But I think the way people think about AI... I define three categories of folks who adopt AI technology. The first bucket, which is usually the place where I see the most traction is, "I am overwhelmed and I need," for lack of a better word, "administrative help. I want to reduce all my low value manual tasks that I want someone to go auto-extract it for me." So that's a pretty strong business case for saying, "I have a massive transaction, I have a divestiture, I'm moving from one platform to another, I have a huge load of documents I have to review in a short period of time." That makes a very simple but compelling use case for AI, which is where a lot of people start.
Teju:
However, the second bucket of people actually think about AI much more than, "I'm going to reduce the administrative burden on my team, and have them do more strategic tasks, and I'm going to let AI replace that lower tier." In that particular case, where I have found traction and people will invest additional resources, particularly legal operations professionals who can actually mine the data for insight, and start to manage risk more effectively, be more proactive in management. So that second category already knows that AI is the way to go. And then they're really thinking about, "How do I prioritize, so that the rollout is pragmatic, and that I'm not biting off too much."
Teju:
The third category are actually what I would call innovators. They're the ones that try and test out newest and latest technologies, they're pushing the envelope in terms of, "Here is my use case, and it's unique and it's different, but I really need it solved. Not because I want administrative burdens reduced, not that it's a people issue, but I can actually be competitively different in the marketplace because of it." And that segment is the one to watch for, I think.
Hemanth:
Amazing, amazing way of putting it. In a way, you put in three categories; walk, run, and fly. There people who just want to get started, and there are people who want to run better and efficient, faster, and there are people who want to do better than anyone else, and then do it faster.
Teju:
Absolutely. And that's where I think, to answer your initial questions of, "Will there be a loss of jobs?" The answer is no. What you will do, will be different.
Hemanth:
That's right. Yeah, in my opinion, Teju, it's a legal assistant, AI will be the legal assistant; do some of the mundane things better than what humans can do, but at the same time, without humans it cannot. It is very important to have that positive relationship with the AI to get the best out of it. And also, Teju, from your perspective, we as vendors, are we doing a decent job or there is a lot of work to be done to ensure, to make it easy and palatable for businesses to adopt AI?
Teju:
There are two distinct camps today. There are what I call the data and analytics providers, and there are many of them out there that do extraction and review. Then there are the platform providers, from the big ones, think about the Gartner Quadrant, to newer nimbler exciting technologies; Malbek, I put in that cool vendor quadrant. Where you're taking and learning from what you have done in the past and where product adoption has failed, to incorporating the newest and latest technologies and tech stacks, so you're not burdened with 15 years of legacy code, but you're able to start fresh, and apply the latest technology, to bring to bear this transformation.
Teju:
What I have found, is the approach that has worked well, is something very similar to what you guys promote. You want to start with a more practical, pragmatic use case that is easily understood by users who are going to use it. Technology is exciting, lawyers can get really, really excited about being able to find 18 points for a termination provision to track automatically, but does it really serve the end user? And I think if you start with the advice and guidance that I hear a lot from platform vendors, I think that's the right approach to say, "There are a lot of things that you can do to streamline your business." You're taking a holistic view of the business. Not every problem has AI as a solution.
Hemanth:
That's right.
Teju:
You can achieve a lot of streamlining and automation, by connecting your contract management system to other systems of engagement. Whether it's a Workday or SAP or Salesforce or what have you. You will get a tremendous amount of lift from connecting various departments across the enterprise. You will get enormous amounts of lift by actually standardizing and harmonizing your templates, and therefore your contract types that you're going to route through for workflow. So that's another automation that people don't think about as a holistic way to streamline, and you get a pretty significant lift out of that too.
Teju:
And then you have analytics tools, where you don't necessarily need AI on the front end, but you just really need intuitive dashboards that can get you the information you need without necessarily applying AI to everything. And I think, with Malbek, if you think about those three things that I've talked about, you've done all three effectively. And the last piece is really applying that machine learning capabilities, because now you have a good purpose of documents, you have analytics, and people understand adoption is there on the core platform. And then you can apply advanced analytics which I call AI, to it.
Hemanth:
Venture capitalists has us the vendors put anything and everything in the AI. Don't try to force it, to what you're saying. Put the AI in the right place where it is helping, but build the foundation first. Build the best foundation, and then build on top of it. So that's what you're saying... Did I get that right, Teju?
Teju:
Correct. And you're saying the same thing too, Hemanth, that we need, first and foremost, a scalable platform. Because what I find, being on the services side... And it's a good business model for services businesses, where every three years you switch out a platform and you start afresh. But it isn't a really good model for the end user, because the cost is not even licensing, cost is not implementation, cost is adoption and change. People, at the end of it, have application fatigue.
Hemanth:
Right, absolutely.
Teju:
And you want to avoid that. So if you're going to provide guidance, you want to be able to say to customers, to say, “have you bought a scalable platform?” And I always liken my contract management solutions to buying a house. If you really want a condo then buy a condo, but don't buy a condo thinking you’re going to turn it into a McMansion, because you will run into space constraints, you're going to have permit issues, you're going to have all of those challenges. So buy a scalable platform knowing where you want to be not for now, but buy for a longer horizon. You can implement what you want just now, you can do that if you buy a scalable platform. That's the distinction, I think, even with AI. You want to buy from a technology vendor that has the capability to scale when you need it but you don't want to invest in a platform that does one or the other. And then you're figuring out integrations, they’re always more expensive.
Hemanth:
By the way, that's great advice that you gave, thank you for that. Switching gears, Teju, one story I want to tell you, is... When was this? Probably nearly about seven, eight years back, I was working for this public company where we had to go through reinstating our financials. And guess what was the bottleneck? The non-standard contract. So with that, company literally spent millions of dollars to identify which were the non-standard contracts, non-standard terms that company committed to, how do you recognize that revenue. They had to go look at all those because they found non-standard ones quite a lot. And auditors flagged it, and they had to do that.
Hemanth:
So now- by the way, that was one of the areas where I started thinking about contracting- so now, fast forward years, now there is a customer who is using our AI solution to find out what's good and bad in M&A contracts. They want to see quickly what are the good and bad. What are the interesting use cases... I mean, I just gave you that M&A as one of the examples. So any other interesting use cases that you have seen, Teju, in the market where AI is being used efficiently?
Teju:
Now, it's interesting you bring up the standard and non-standard contracts. I'll give you a couple of examples that come to mind. One of the customers that we had worked with recently, they had actually done all of their tracking, being forward thinking, to say, “I actually want to know which of my contracts have non-standard versus standard provisions.” And they went through and tagged a number of contracts, a pretty significant amount of their contracts, and went through in great amount of detail to do that. Now, that was all wonderful and great.
Teju:
Two years later, they actually updated their standard template. So what was standard in 2015 is no longer standard. And so, always happens to businesses because regulatory requirements change. I bet if you were to look at standard contracts, or force measure in 2021, they would be very different from what they were in 2018. And so one of the areas where AI is enormously helpful is, if you didn't have the AI capability, you wouldn't be able to go update 5000 contracts with the click of a button to update it to this new standard. You'd have to go and do that manually. So that's where I think AI is phenomenally helpful in going back through an entire purpose.
Teju:
Another example that I came across recently was again, tagging. Where people think about termination provisions. And people think about termination in a very big, buckety way. And so people said, “Okay, I want to know if there's a termination provision or not.” And that's all they care about. And most people when they think about a termination provision, they really think about termination for convenience, they assume cause is always going to be there.
Teju:
Now, what ended up happening for this particular customer is, they wanted to check whether there are early penalties, or early termination penalties. Now, if you want to basically use a manual method of going back and checking your contracts, you're going to have to do that for all the contracts that you said yes to termination. But if you were to use AI, you could actually just go tag 10 and say, "This is what I'm looking for. Well, find me every contract that has a early termination penalty clause and update that." So that is another example where we are seeing a lot of use of what I call human insight coupled with AI intelligence.
Hemanth:
Very practical and very useful use case that you gave, Teju. We have seen use cases like this, like customers specifically looking for the standard and non-standard provisions using AI. And as you said, the business keeps changing, and you may have seven different types of non standard provision. So can you train your AI to say, “Okay, these are the seven, if you see this, then tag it as X1, and if you see this, tag it as X2 and so on." And quickly able to go see those and also work update, like you said, mass update. That's an amazing use case. So, Teju, this is been an exciting discussion, very, very thoughtful discussion. Any closing thoughts from you, Teju, on this topic?
Teju:
If I focus only on extraction for now, I think humans, we could devote an entire section to AI-assisted contract review, and redlining, that's a whole other topic in itself. But if you were to think about extraction, manual versus AI review, so to speak, the way I think about manual versus AI review in terms of three categories; one is accuracy. One is, if you can train the AI effectively, most platforms are pretty accurate now in terms of, if you do a human versus machine, you train them properly, the machine will be right a significantly higher percentage of the time than humans. Not because humans get it wrong, humans don't apply the same rules consistently, whereas AI will apply consistently.
Teju:
In terms of flexibility and speed, again, AI can win if you have very clear instructions on what AI should do. So multiple reviews, can be done faster. Now, where I feel that the AI has limited functionality today, and this is where I think a lot of innovative teams are working towards is that, today's AI looks at a single document to extract. So if I have a master services agreement, and I have five amendments, it doesn't extract from the fifth amendment to update. That is one limitation.
Teju:
The second one tends to be around nuance and summarization and contextual review. Total amount is another example. We've encountered a situation where people want to know, "What's the total amount of this contract?" And you would think it's pretty straightforward, but it isn't, you could have 15 SOWs and change orders underneath them.
Hemanth:
That's right.
Teju:
And so it's hard to summarize and quantify that automatically. And so those are two things that I see that I think that when people think about AI, it's not a panacea for everything, you still need governance, you need data governance around it, you need to have the ability to see how it will push information to different systems, and how it will ingest information from different systems. And does the AI have that capability? And last is how trained is the operator going to be? I don't want to be responsible for a really sharp steak knife if I don't know how to cut.
Hemanth:
Absolutely.
Teju:
And that's what AI can do, you can go rogue with it and have a completely poorly tagged system, because someone didn't take the time to train it properly.
Hemanth:
Absolutely. And that's a great, great way of putting it, Teju. It's so true, actually, you can go over the top and not get the value out of the AI. Be pragmatic, understand that there's a human element to it, also. Like you said, governance, don't assume everything is magic. But at the same time, try to do the mundane tasks through the AI and see the vendor that you are selecting, the platform is capable of doing many things, and also it can grow.
Hemanth:
Those are some of the high-level items that I took from the discussion. Again, Teju, thank you for the wonderful, wonderful discussion, very thoughtful, and great ideas with this list. And looking forward to many other discussions like this, Teju.
Teju:
Likewise, Hemanth. And I'm looking forward to all the exciting developments coming out of Malbek. I'm excited.
