EPISODE 204 – FAST-TRACKING CONTRACT ANALYSIS WITH LEGAL SIFTER’S KEVIN MILLER

If you’re doing heavy contract drafting, review, editing, and/or e-discovery, then Legal Sifter can help you understand lengthy and complex contracts within minutes using their artificial intelligence and machine learning technology.

Legal Sifter uses natural language processing (NPL) & machine learning, to process unstructured raw terms, conditions & words into well-structured data & insights ready to be used to make quick & well informed decisions, saving both time as well as money for both attorneys and anyone facing the reading and interpretation of one or more legal agreements.

Kevin Miller, Chief Executive Officer of Legal Sifter explain how their contract interpretation artificial intelligence software helps the lay person and experienced attorney both read and understand contracts faster. The software gives you a second set of eyes to find the hidden terms & conditions that may be overlooked in a contract.

The legal industry is being transformed by artificial intelligence, and democratizing tools like Legal Sifter are allowing people who may find it difficult or too time consuming to hire an attorney to interpret contracts and make education decisions.

 

Podcast: Play in new window | Download

Hey, this is Richard Jacobs with Speakeasy Marketing’s attorney marketing podcast and I’m speaking to Kevin Miller, CEO of Legal Sifter. It’s an application used to help people understand contracts which can be quite complex at times.

Richard Jacobs: So welcome Kevin, how are you doing?

Kevin Miller: I’m doing great Richard. Thank you so much.

Richard Jacobs: Thanks for being on the podcast. Really appreciate it. I gave a 2 second description. Can you give a better description of what Legal Sifter does and how it works?

Kevin Miller: We are an artificial intelligence company. We are based in Pittsburgh, Pennsylvania and born originally in Carnegie Mellon University and we are using a form of artificial intelligence specifically machine learning and natural language processing to help people digest contracts. The way it works is customers upload contracts that have been executed or are in draft form. We have built models that we call sifters that when clicked and selected, the customer can navigate around the contract faster, they can zip around the contract faster. They uncover hidden terms and conditions that they may otherwise miss when they are reading through a 20, 30, 40 page single space document. It helps to ensure that they don’t miss those things by highlighting the clauses that might relate to indemnification and finally the sifters help the customers to organize all of their data to structured fields that can then be exported to Excel and other databases for analysis and reporting.

Richard Jacobs: So, for the average person that doesn’t know much about artificial intelligence. There are all kinds of crazy things that people think. Can you break down what are the elements that make your system intelligent? Maybe start with the natural language processing. How does it work and how does it understand the document?

Kevin Miller: I think there are a couple of major concepts to understand when we talk about this technology. The first concept is that it’s math. It’s really fancy math. It’s machine learning algorithms and natural language processing. That’s the jargon and in the middle of that is the word algorithm. So think about, I always talk about it as extremely fancy regressions for those of you who took statistics but math problems that are predicting outcomes based on input data. So the first concept is that it’s math and it’s using forecasting to represent results to the user. The second concept is natural language processing. So it’s applying that math to language. In this case, text in a document. What it’s doing is it’s looking for sequences of letters and numbers and their relationship to other letters and numbers to say that that sequence of letters and numbers is probably governing law. The reason we think that it’s governing law or the reason we think it’s indemnification or the reason we think it’s a start date in a contract is because we’ve trained on and built mathematical models to see unique sequences of letters and numbers and interpret those as a concept. So we are applying math to letters and numbers and language. That’s the second concept. The last concept is learning, the machine learning. The great thing about this technology or the thing that makes it different from the software that you and I have been staring at for the last couple of decades is it learns, meaning if we feed it more information and more examples, say of governing law or start date in unique sequences of letters or numbers, it can then improve the mathematical model such that it sees the other unique sets of letters and numbers that perhaps it has never seen before, we can infer that it is part of a set called governing law and that’s what’s cool because if you think about how people navigate through contracts today, they have to generally scroll through the pages or flip through them to find the concepts they are looking for. Now, sometimes in the document, you can apply keyword searches and search for certain keywords but if you don’t know the right keywords to search for or people use different words, you can easily miss those concepts. With this type of technology, math on top of text that gets smarter over time and recognizes unique series of letters and numbers, it can see unique phrases and infer that it’s a part of a certain concept and highlight that to the user. That’s important when you are reading a contract and one of the things we all read a contract for are things that can get us in trouble. We want to make sure you don’t miss anything. So these sifters are like someone looking over your shoulder on top of your document to help you navigate these complex things and make sure you don’t miss anything.

Richard Jacobs: So the natural language processing, does it essentially build up a library of mathematical concepts that apply to legal concepts and then when you present it in a new document, does it compare that against the library of concepts and also going a step further and being able to infer that okay, this looks similar to that mathematically so it’s probably included in that. Does that make sense? Is that how it works?

Kevin Miller: Yes. That is a pretty good description of how it works. Yes.

Richard Jacobs: I understand. Very strange. I don’t know if you can do this but is there a way to show how it would understand one concept in a document. Let’s say that something.. Is there a really simple element in a contract that the machine can understand and can use. How would it understand that? How would it mess up and how would it correct itself?

Kevin Miller: Today we don’t ask the machine to do anything with what it finds other than to highlight it for the user. So it’s not trying to replace a person. We are trying to supplement the person with the technology today so we have these major problems. I was just at a conference, The International Association of Commercial and Contract managers which was out in San Diego which is an organization of contract management professionals consisting of lawyers and non-lawyers, 40,000 people strong. They presented the results of their annual benchmarking survey and they said that only 12% of their population agreed or strongly agreed with the statement that contracts were understandable or easy to read or something to that effect which said differently means 88% of the world find contracts difficult to get through and I think it’s probably higher than that. I think that our first goal with our first set of products that we’ve commercialized has been to use this technology just to help the people navigate through the contracts and just to help them organize the data inside the contracts. So often times people want to know things like when is my contract up for renewal? The way it happens today is people have to read through those contracts and write that information down or maybe they want to understand which of my contracts are assignable if there is a change of control or if we get bought. Somebody has to read that document and write that stuff down. We provide a tool that helps those sort of processes go faster. So that’s how we are using it today for navigation and helping to ensure that people don’t miss anything in organizing their data. However 2 years ago, we built a product that did something more than that. We built something called the Freelancer Scorecard and we are sitting on this technology right now and you will see us come out with something like this in the coming quarters. We use this technology not just to recognize the concepts inside of a contract but also to give the user feedback on what is found inside the contract in context of the contract itself. We targeted freelance software developers. These are folks who do 10-99 or contract work for software development or web development work and they oftentimes sign contracts with customers that they never show to a lawyer because they can’t afford the average cost of an attorney. So they sign a contract, they do their work but they don’t get paid because they signed a bad contract. Our team built a solution called the Freelancer Scorecard, you uploaded your contract if you were a freelance software developer, you upload a contract you were negotiating and our software used machine learning and natural language processing to “read it” and then to provide after a few seconds or a second or two, a bunch of in context feedback next to the clauses and offered help and support and suggestions on how to protect themselves, themselves being the Freelancer Scorecard or freelance software developer. So we told them, hey look. You need to look out for this, you need to look out for this and you should consider adding this or subtracting this from the contract. It was launched to the web in August of 14 and it got 4000 customers within a couple of weeks. We are sitting on this technology right now. We are the first people in the world to try to use NLP, Natural Language Processing and machine learning to give that kind of in context feedback within a contract. So you are going to see us go back to that in 2017. So when you talk about how would this tool or how would these algorithms recognize “a concept” and do something with it. That’s what we expect to do. We expect to partner with the legal community and launch a series of tools that allow the users to upload these complex documents that they are negotiating or have negotiated and get an automatic in context read on what’s inside of it which is so much of a battle with contracts given their complexity.

Richard Jacobs: I would think that when you look at a particular niche, employment contracts. Now the thing is you’ve looked at thousands of them, 90% of the clauses would be boiler plate and the same and then there would be a few outliners. So I would guess that your software would know the library and what’s common after having been trained on X number of contracts and that’s how it could, not advise you but point out, hey you have this clause that occurs very rarely or never seen this before or you are missing this other one that occurs in 90% of contracts. You should have it. Is that how it works?

Kevin Miller: That’s right. The spirit of what you are talking about is correct. This math, the artificial intelligence in today’s world. It’s not generally intelligent. It’s actually quite dumb but it can be trained for a very specific purpose and then it can be quite efficient, quite intelligent and quite helpful and most importantly quite fast. I like to liken today’s technology to a 4 or 5 year old child. I have children, I have teenagers so I’ve seen this progression and you can teach a 4 or 5 year old child to put his or her dishes away after they eat dinner or you can teach a 4 or 5 year old child to clean up their room. You could teach them to do a specific thing but they are unlikely to teach themselves and they are unlikely to do it on their own and not without a lot of instruction. But then in a few years they are 8 or 9 years old and then a few years after that, they are 15 or 16 year old and they can do other things, teach themselves and start to advance quite quickly. That’s how this technology is going to progress. What it means is if we trained these models to look for cob salad or the recipe for cob salad and the recipe happened to be in a contract, it would find it. The flipside is if we train it to look for governing law and find it in a restaurant menu, it would find it but until we tell it what to do with it, it’s very dumb. The nice thing about many contracts is that they do follow certain patterns and so it’s somewhat easier although this is really some hard stuff, somewhat easier, this is a good place to begin to apply the technology to text because you are seeing a series of terms and conditions that you are seeing over and over again. So you can actually train that 4 or 5 year old to grow up and help you out.

Richard Jacobs: What kinds of contracts has the system been trained on? Is it good enough where if you showed a genre of contract that it hasn’t seen before, will it still be able to interpret it or it has to be pretty narrow based on what it’s been trained on?

Kevin Miller: We’ve trained our sifters on terms and conditions and so think of a contract not as one thing but as a stack of terms and conditions even that if a term or a condition appears in a contract, it doesn’t matter what the format of a contract is, it will find it. For example, start date or termination date or renewal clause or governing law or jurisdiction or assignment, these are clauses and terms that appear in many different types of contracts. Therefore the technology can be applied to those contracts to the extent that you care about those terms and conditions. We have seen everything from employment contracts to procurement. We got it from the seller’s side and from the buyer’s side, general commercial contracts, we’ve seen financing documents, insurance documents, non-disclosure agreements, it’s a fairly long list of agreements that we’ve seen because our customers have run the gamut. We’ve sold this to general counsels, to procurement, to sales, to operations, to law firms and the reason for that is all of those people have different viewpoints of looking at contracts. Because we are solving the problem at the clause level and not necessarily the contract level, we’ve been able to apply our math in many different contexts.

Richard Jacobs: That’s great. Okay. Any types of statements or agreements that the computer is having a hard time figuring out or you are having a hard time mathematically modelling? Where does it get murky?

Kevin Miller: I think when you start to get more precise. So right now, we are using the technology to recall concepts, so find stuff that’s probably in the realm of indemnification or assignment to the extent that you want us to go inside the assignment clause and tell you what type of assignment clause it is. Assignment clauses usually fall into a couple of buckets, I can freely assign this without your permission, I have to assign it with notice, I have to assign it with your consent, I can’t consign it or it’s one sided, one party can assign but another can’t and this is in the event, say it gets bought by a company and then there are all kinds of exceptions. You can model all the different assignment types but to actually have the algorithms then recognize that this is Type B of the 15 different types of assignment clauses and then do something with that, that requires another level of sophistication because you have to recognize that it is in a set of assignments and this is how I think about it or math guys may explain it. You have to recognize that this is in the genre of assignment and you have to recognize that this is an assignment clause that’s saying a specific thing. So what that means is we are able to recall what this clause is in assignment. We are not yet able to say in a precise way, this is always something in this type of assignment clause, therefore we should do this. We can’t be as precise with our read on that yet. That will come in time, we just need more time, more investment, RND etc. That’s for the level of precision, the more precise you want to be the more effort we have to get to. So it’s like a lot of things, you can get to 95% or 98% but getting that last 2 or 5 percent often times makes the difference on another concept or another offering or another value proposition and there is a lot of effort that goes into getting that 2 or 5 percent. That’s how I would explain it, I don’t know if I did a good job but that’s kind of where we are struggling.

Richard Jacobs: It made sense. It makes me think of the question, if your system is trained on 3 different types of contracts, does it hurt the system or dilute it? It keeps learning more and more things or is it each training data set discrete and kept on its own or does the system help itself get better in interpreting language?

Kevin Miller: On a broad level, as we improve a model, as we get more experience, we are re-training our models and then redeploying those models across all our customers where that’s appropriate. So, to the extent that our models are performing at a certain level today, we just did this last week. Last week we realized a bunch of improvements to a handful of models and a bunch of improvements to a handful of sifters and then we released the new sifters. All of that was a function of new training data, in many cases, new focus and we refined the input to improve the math and then we redeployed that on our customer site. We are doing that on an ongoing basis but in an ad-hoc manual way, so we ask our customers that they give us access to the contracts for sifting purposes. We ask them for permission to use their contracts for research and development purposes. To date, all of our customers have said yes because they’d like our models to get better and better. Eventually what will happen and eventually could mean inside of 12 months, not quite sure, we will set up the system so that the feedback mechanism will be more real time such that our users give feedback on the quality on the lack of quality of a particular, from their perspective position. We get that feedback from the user communities and we improve the models that much faster so there will be more of a real feedback mechanism that will be built into our tools over time that will improve the intelligence. Finally, we will also do this in parallel, we will also teach the software at some level to go out and train itself. That’s where a lot of the magic happens. That’s a much more complex thought, there is a lot of trickiness to that, it’s not in any way, shape or form easy but it’s the path that we are on.

Richard Jacobs: What do you think is realistic and what do you think is fantasy over the next two years? What is possible and what’s say really way out there kind of science fiction and not likely to happen for quite a long time?

Kevin Miller: That’s a really good question. Here is what I think, here is what is really possible. Really possible and only because we’ve done it before. We did it in 14, it was really a beta format with the Freelancer’s Scorecard and what we are going to launch next year. What is really possible is for these algorithms to present the best practice point of view of a company internal to itself in the context of the contract. So what does all that jargon mean? A lot of times when you deal with an organization and let’s take a data person contract and you’ve got a bunch of buyers all over the world that are buying stuff and they are not lawyers. Before they can sign off on something that they are buying but they’ve got the customer’s paper or the supplier’s paper or theirs, the person with the contract has to get to an attorney. Typically, like a lot of personal contracts, a lot of this stuff looks the same and if the buyer were better trained, he or she and you typically see this, I did 6 years in supply chain, so I know that the more experienced buyers can take on more and more responsibility for the negotiation. They’ll know what to do with so many indemnification clauses and the governing law clause and the payment term clause and they won’t have to go to legal but when you have a new buyer, you have a buyer that doesn’t understand that stuff, you always have to run the contract back and forth to legal and it delays everything. Wouldn’t it be great to put the general counsel’s mind, put her mind in the context of the document that the buyer is working with in real time? That’s possible. The idea that the buyer could upload a draft of that supply contract whether it’s the customer’s or the supplier’s paper or theirs and the give it to read to the tool and put down the 25 things in context that the buyer needs to think about such that the buyer can make his or her red lines in real time. Hopefully they will not have to go to legal to get the deal done and look out for the gauchos to make sure that they don’t sign anything they shouldn’t. That’s the first step, that’s going to happen, that’s within reach. I’m telling you that’s within 6 months’ reach, not within 6 years’ reach. What’s not going to happen in two years, so I think that’s what’s going to happen. You are going to see these playbooks played out. It’s almost like your grammar check and your spell check in Word except that the check that you are going to get is your general counsel’s brain expressed through these algorithms and expressed through our technology and others. What’s not going to happen and what is still years away and I don’t really know what the timeframe is, 4, 5, 6 years. You upload a contract and the tool “reads” it all and gives you a summary feedback that says, Hey look Richard, you don’t have to worry about this non-disclosure agreement. You can just sign it or Hey Richard, I’ve read this whole thing and I know how you think about things and you can probably let these 5 clauses go and just kind of go ahead because we know that we don’t care too much about this and we only care about this one clause. You need to go and get this residuals clause out of this non-disclosure agreement. That’s a wiser way. That’s a level of complexity and a level of conclusions on top of conclusions that are hard to model today but will be within reach in years. It’s a real dream in the short term and maybe it’s five years out or maybe it’s 7 years out, I don’t know but it’s a ways away for us to be able to do that and provide that kind of, people want to see this on contracts that actually work. You are going to see us get there incrementally.

Richard Jacobs: Outside of contracts, I don’t know if you are in contact with any other AI involved applications, that article in the Guardian, I think, about a month ago where this team fed a machine learning system human rights cases in the EU and then asked it to make determinations on what the outcome would be on more cases and they said that they got like a 79% correct ruling rate. They call it like an AI judge. Anything else out there that you’ve seen from other companies or other people that is going to be interesting in the AI space of law?

Kevin Miller: Well, I think what this Ross company is doing with legal research is interesting on top of IBM’s Watson platform, I think. I don’t know if you know this, Ross company, rossintelligence.com. I think you guys may have done an interview with them.

Richard Jacobs: Yes we just recently talked to them.

Kevin Miller: I think what they are doing on top of legal research, I think Watson is well suited for that type of work and I don’t know the folks but it sounds like they can really harness that and are putting it to good use and I think they are going to get a lot of looks and a lot of interest. So I think that’s one area where you are going to see this stuff. I think another area where you are going to see these applications of machine learning is the general challenge of reviewing emails or lots and lots of documents for pertinent information from a risk perspective or even an e-discovery perspective. E-discovery, those firms have been using some machine learning for a while in the form of predictive coding. They are using it at a high level saying hey is this document relevant or not. Yes, no. I think we’ll see quite a bit of advancement in that approach in the coming years from those companies and from companies that might be better than those companies that are doing that type of stuff because when you talk to lawyers about predictive coding, they’ll tell you that it still has a long way to go but it’s promising in the E-discovery space but I think you’ll see more and more of a focus on Hey can you go into all of my emails and tell me if I’ve got some sort of a hidden liability there based on some promise that somebody made over the course of last year? It’s kind of a Hillary Clinton story, you know, Hillary Clinton had all those emails and the FBI comes out and says we have to get through all these emails and in the legal system, we thought gosh, we are not too far away in able to being help out with a problem like that. I’m not sure how they got through all those emails but I’m sure they used some technology that’s not too far afield from what we are working on because the ability to process the text and identify risks quickly, that’s extremely valuable in a world where the compounding creation of information is just overwhelming, I suppose. So I think those are some areas where I think you’ll see this legal research space with folks like Ross. I think you’ll see further advancements in predictive coding and E–discovery, specifically litigation and then I think you’ll see some early Tsunami warning systems on things like email inside of people’s companies as general counsels try to get a handle on the exposure that companies have. Those are some of the things. Otherwise I think you will see us and maybe a couple of others try to do what we are trying to do which is we want to bring affordable legal services to the world by empowering people with artificial intelligence. Lawyers, in general have not been able to scale themselves and it is one of the primary reasons why they are so expensive and that’s because they have not had technology that gets at the heart of what they do every day. It’s not the lawyer’s fault. This technology helps them read things, helps them analyze things, helps them make decisions and recommendations which is what lawyers do. I heard that from Thompson –Reuters, I heard that lawyers read things, they analyze things so I thought that it’s a very good model. They haven’t had anything to lessen scale and I think what we are doing is going to help them scale and that will eventually bring down the cost of legal services and hopefully help close a lot of the justice captives out there and actually double the market. You know that legal services is a $750 billion, $850 billion market by some estimates and it should be twice that because the unmet demand is just enormous and the only reason that the demand goes unmet is that the lawyers can’t work fast enough and at a low enough cost to serve that customer base and as a result, a lot of businesses and people can’t get to where they need to get to.

Richard Jacobs: One thing that came to mind is nowadays if regular people were confronted with terms and conditions constantly, when you try to launch an app, when you use a website, there are privacy policies and all kinds of things. Have you thought about an incarnation of your system that helps people when they are confronted and they have to make a decision in 30 seconds whether to use an app or not or some other service, GPS and they have this 30 page terms and conditions type of document. Do you have your system distill that right away maybe live as they are looking at it and tell them of any warnings if they are opting into something? I’m going to give you a quick example. Google Drive, I don’t know if you know this but part of their terms and conditions is that they are allowed to make derivative copies of anything you put on Google Drive. I don’t know if you knew that. It’s just pretty extensive rights that you are giving away when you opt in to something as simple as that or let’s say when I use Google Navigation on my phone, I mean their terms and conditions for use are pretty extensive and it could cause you liability and people might want to be aware of that. So I don’t know if you’ve thought about an incarnation like this to help people because when you are in a time crunch and you want to use something you are not going to just sit there for 20 minutes and read it especially if it gets updated every few months.

Kevin Miller: Yeah, so I think what you are talking about is a natural extension of what this company built with the Freelancer Scorecard in 2014 and we absolutely intend to go in that direction. When the vast majority of the world is tied together. Arguably the most important document in Global commerce is the contract and it is the least understood of all the documents. It’s so painful and as I stated before, 88% of the profession dedicated to looking at contracts all day find them difficult to read or work with. So the average layperson doesn’t have a chance. A template is not going to help you with that and a lawyer can’t help you with that on his or her own because they are so expensive the way they are doing business today but we perceive a world where what you just said happens in seconds supported by the legal community, enabled by this type of technology and it’s a natural extension of what we built 2 years ago. As the math gets more precise, as the processing speeds improve, as deep learning becomes more pervasive, you will see us and others aggressively pursue that type of consumer and business applications because it’s on our mind.

Richard Jacobs: Okay, very good. One last question and I’ve been asking several AI companies on this. What’s your thoughts on this? Is it possible that a system which would review a thousand divorce cases in the state of Minnesota and then a family law attorney would use it. They get a new divorce case, they feed in the facts of the case to the machine learning program and the machine learning program acts like a recommendation engine and tells them to go this way in a case, there is a high likelihood of this happening, if you go this way, no. Is it viable or is it fantasy? What do you think?

Kevin Miller: I think it’s completely viable. Machine learning and I think Watson is a good example of this, is well suited to digest a whole bunch of information and make predictions based on the past. What you are talking about fits into that simple statement and simple hypothesis. So, to the extent that someone is able to collect the information and find the talent. You don’t want to underestimate this, you can’t go and just hire a developer to do this, you have to go to school for this stuff. It’s math and computer sciences. This stuff is really hard. I was told recently that some of the stuff that we are doing and I was in Pittsburgh, it’s the land of driverless cars. So we’ve got Uber driverless cars all over the place in our city and the heart of robotics and the heart of machine learning, currently known as the number one machine learning school in the world with Stanford and so we are at the heart of this revolution that’s happening and I was told recently that what we are trying to do in natural language processing is in some ways harder that what they are trying to do in making a driverless car. So you don’t want to underestimate the difficulty of what our guys and other companies like us are trying to do. That being said, to the extent that somebody can get the information to a person who can do the math is possible.

Richard Jacobs: Okay. Very Good. Last question or two. Who do you want to appeal to on this podcast and give your contact information to for them to possibly become a customer for your company? Is it attorneys? Is it individuals? What service can you provide to them now or in the near future that they may be interested in?

Kevin Miller: I think if you are a general counsel of a small, medium or large enterprise. We’ve had Global 100 companies all the way down to 14, 15 people, person organizations work with us to date. You are a general counsel to start with if you are operations, or sales or a law firm, with our current suite of products, you are in the second tier of customers. Here is what we can help you with. First, if you have a stack of contracts and you need to know somethings or many things about what is in that stack of contracts, we can do that for you. We have something called the Contract Sifter Service where we use our AI in conjunction with service providers on multiple levels including on onshore attorneys to sift out the terms and conditions and the data that you want in a structured format. We could do that for you or we could provide you with the tool, contract sifter and you could do it yourself faster, cheaper and better than you could do it on your own with higher quality because a person plus an algorithm is stronger than either by itself. So if you are a general counsel or you are a procurement person and you are concerned about the assignability of a bunch of contracts because you just got bought or you just bought somebody and you want to re-assign all those contracts to your company, someone needs to read all those contracts and understand what is going to happen. Do we need to notify people or can we do it with consent or can we not assign these contracts in order to move forward? If you want to know things like when are all of the end dates in my contract so I can manage my renewals, somebody has to read through those contracts. We can help you do that faster, cheaper and better. We had a project for a large multinational bank, one that you would know, one of the largest ones. They had 31000 procurement contracts that they wanted to go through to understand whether or not they were impacted by the EU safe harbor ruling that happened last year where the EU safe harbor policy has governed how we handle data across the ocean between Europe and United States and other countries for years. It was invalidated in 2-4 of last year. So this large bank needed to know which of these 31000 contracts were impacted by the safe harbor ruling. We did that in a weekend. There is no way that a human could have read 31,000 contracts and figured that out without our technology. So, a lot of times general counsels are presented with this challenge. Hey, we just bought this bank and they use off-shore suppliers. I know we have customers that don’t allow us to use off-shore suppliers. Which customers are those? Nobody can tell you those things without reading the contract. So those are examples, procurement has these cases, general counsels typically in due diligence situations, they have to read and digest contracts quickly in a short period of time. We have tools in our service that can help them do that. On the sales and operations side, it’s similar but it’s different. Operationally, if you have a company where you are driven by contracts, where commercial business doesn’t start until you have a signed contract with a customer. Typically what happens is when that contract gets signed, it gets read multiple times throughout the organization. It’s an extraction process. So it sits right there in order management and it gets passed around to marketing because they care about publicity and then it goes to finance for payment terms and then it goes to service and manufacturing or executive leadership or sales compensation and people read these contracts 4 or 5 times. Wouldn’t it be great if it got read once because it got read for the same things every time, the same data was extracted every time and then that data was elegantly pushed out to everyone quickly throughout the organization. Costs would go down, quality problems would lower. The cycle times to celebration would go down. We think there is a nice use case there at the point of order management. Then on the sales side, there are some businesses who are outsourcing procurement for a company and in their sales cycle, they need to review contracts just to put together a proposal. We have a use case right now for one of our customers with that. Or if I’m a sales person, I want to know what have I sold to these customers? Are there any triggers like some sort of payment accelerator if I hit a threshold? What are my renewal clauses looking like? How do I manage my renewals? I need to get a bunch of information out to manage my sales operations. If you are doing that by hand or if you are doing that manually, leasing companies or real estate companies are like this. They are managing lots and lots of commercial leases and they have armies of people that are reading these documents. Those people would be faster, cheaper and have higher quality results if they were using contract sifter. These are the types of customers that we are trying to target with Contract Sifter. Last point, law firms who are interested in becoming able with artificial intelligence and turning their service only offerings into AI enabled service offerings, they need to come talk to us because we are going to come out with a series of solutions that we hope to partner with the legal community, that will enable them to once and for all scale. Lawyers generally can’t make money while they are asleep because they don’t have anything generally that allows customers to interact with them or interact with their brains unless they are awake and they are billing and they are doing business. That’s not the way it is. So many other businesses. We can for the first time in history provide them with a suite of tools that are coming out that will allow them to scale and we want to do that with them. So if there is a law firm out there that is interested in getting on an AI journey because this is going to be bigger than the internet and computers and mobile and cloud and it’s going to change everything. If they are interested in getting on that journey, we want them to call Legal Sifter.

Richard Jacobs: Okay. Very good. Literally, what’s your contact information? Website, phone number etc. email?

Kevin Miller: My name is Kevin Miller. I serve as the CEO of Legal Sifter and we are www. Legalsifter.com and you can contact me at kevin@legalsifter.com or my phone (412) 370-1356. You can also contact Legal Sifter at our main number and talk to any number of our people at (724) 224-7348 which is (721) 224 –SIFT and we’d like to hear from you.

Richard Jacobs: Well, very good. It’s been a great conversation, a lot of useful info and I think attorneys listening, people interested in AI and it’s applications will all get the benefit from understanding it. You boiled it down really well. So thanks again for your time. I really appreciate it.

Kevin Miller: Thank you, Richard. We look forward to the podcast and thank you so much for your interest in Legal Sifter.

 

Podcast: Play in new window | Download