(Credit: Zapp2Photo/Shutterstock.com)

Of the many challenges facing in-house attorneys in M&A or a company restructuring, reviewing and organizing company contracts is perhaps the most labor-intensive task. With advancements in artificial intelligence (AI), however, many legal teams been able to reduce the amount of time and effort such review can entail. But AI is far from a silver bullet.

Almost all AI contract deployments in legal relies on machine learning, which allows a program to be taught how to understand a dataset with an almost human level of comprehension. This training, though, is far from automatic or simple.

Just ask who were tasked with deploying an AI contract review project for the legal department at Baxter Healthcare Corp. Key stakeholders in this project discussed their experience firsthand at the “Meaningful Law Department Metrics: Contract Analytics and AI in Today’s Law Departments” session at the Association of Corporate Counsel’s (ACC) Legal Operations Conference.

A few years ago, Baxter’s legal department at faced a difficult demand. “In 2014, our CEO announced we were splitting [the company] in two,” recalled Aaron Van Nice, director of legal operations at Baxter. “Immediately, I was asked to review the potentially 100,000 contracts out there in the multiple countries we do business in.”

Van Nice’s team decided it was best to tackle these contracts by splitting them “into categories: customer contracts, strategic contracts and everything else.” But the company didn’t want rely on outside counsel to identify, review and categorize each contract because of the high cost that would entail.

So they sent out a request for proposal (RFP) for legal outsourcing providers who could handle the task, and they were soon approached by legal service provider Elevate, which whom they had a previous relationship.

Pratik Patel, vice president of legal business solutions at Elevate Services, recalled that Baxter’s team wanted to keep costs down by using a pay-per-document fee structure. “And the only way we could do that was by leveraging technology.”

The task before Elevate, who was ultimately chosen by Baxter, was a multifaceted one. It had to identify relevant contracts in the batch of 100,000 contracts, pull out actionable and relevant clauses in high-risk contracts and extract contact information in lower level contracts so that the legal department could inform their partners of the restructure.

Leveraging Kira Systems’ AI review engine, Elevate first assessed how many relevant contracts they were dealing with. Though they started off with an estimate 100,000 contracts, Elevate found that only 10,445 were “qualified contract documents,” while others were “not applicable” to the project, Van Nice said.

The next challenge was to extract relevant information, such as contact data or clauses that related to certain risks, from each contract. Using machine learning system could automate this identification— but first the technology had to be trained.

Patel recalled that the first time the team ran a batch of contracts through Kira’s AI engine, it could not identify or extract nine out of the 10 data points they needed. But with the next batch of contracts, the solution “found more information.” And after subsequent iterations, which helped to better train the AI, “we found it 98 percent accurate,” he said.

Training the platform, however, was not just a question of allowing Kira’s engine to continuously review batch after batch of contracts. The technology had to be told what was right and what was wrong in its identification of key data points.

To that end, Elevate deployed a team of lawyers who were familiar with the types and foreign languages of the agreements Baxter had to review. These attorneys continuously validated or corrected “the system as it was extracting the information,” Patel said.

Another speaker at the session—Sowmyan Ranganathan, senior director of legal operations at AbbVie—noted that this rigorous and extensive process was one of the reasons it’s difficult for many companies to train machine learning systems AI on their own. “They’re not going to have those kinds of resources that are needed,” he said.

To be sure, training AI with contracts is much easier when an organization already knows what its standard contracts encompass and which contracts it holds deviate from the norm. With that information at hand, its AI system can be more easily trained to understand the baseline of which contracts to work off.

But uncovering deviations within a contract database from the start is an onerous task. Patel noted that a company “can force [their] lawyers and business stakeholders” to edit contracts within a contract management platform. This way, the platform can track and provide metrics on the amount and type of changes being made over a large body of contracts. Yet such a process largely depends on many different in-house teams, whose adherence to the process may not be reliable as needed.

Far simpler, Patel said, was to review a contract for deviations when it first enters the organization. “What we typically see is at the point of storage, you have a contract team that says we’re going to load a contract and check for deviations,” he said.

Yet even while Baxter did not do that step ahead of time, the process using AI, even though it required extensive training, was still more cost effective than other options.

Van Nice, for example, noted in the end, the contract review project was around $500,000 dollars less expensive than what “other LPO providers that were going to do it manually [were offering],” and the AI saved an estimated 5,000 hours of work in total.

Copyright Legaltech News. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.

Of the many challenges facing in-house attorneys in M&A or a company restructuring, reviewing and organizing company contracts is perhaps the most labor-intensive task. With advancements in artificial intelligence (AI), however, many legal teams been able to reduce the amount of time and effort such review can entail. But AI is far from a silver bullet.

Almost all AI contract deployments in legal relies on machine learning, which allows a program to be taught how to understand a dataset with an almost human level of comprehension. This training, though, is far from automatic or simple.

Just ask who were tasked with deploying an AI contract review project for the legal department at Baxter Healthcare Corp. Key stakeholders in this project discussed their experience firsthand at the “Meaningful Law Department Metrics: Contract Analytics and AI in Today’s Law Departments” session at the Association of Corporate Counsel’s (ACC) Legal Operations Conference.

A few years ago, Baxter’s legal department at faced a difficult demand. “In 2014, our CEO announced we were splitting [the company] in two,” recalled Aaron Van Nice, director of legal operations at Baxter. “Immediately, I was asked to review the potentially 100,000 contracts out there in the multiple countries we do business in.”

Van Nice’s team decided it was best to tackle these contracts by splitting them “into categories: customer contracts, strategic contracts and everything else.” But the company didn’t want rely on outside counsel to identify, review and categorize each contract because of the high cost that would entail.

So they sent out a request for proposal (RFP) for legal outsourcing providers who could handle the task, and they were soon approached by legal service provider Elevate, which whom they had a previous relationship.

Pratik Patel, vice president of legal business solutions at Elevate Services, recalled that Baxter’s team wanted to keep costs down by using a pay-per-document fee structure. “And the only way we could do that was by leveraging technology.”

The task before Elevate, who was ultimately chosen by Baxter, was a multifaceted one. It had to identify relevant contracts in the batch of 100,000 contracts, pull out actionable and relevant clauses in high-risk contracts and extract contact information in lower level contracts so that the legal department could inform their partners of the restructure.

Leveraging Kira Systems’ AI review engine, Elevate first assessed how many relevant contracts they were dealing with. Though they started off with an estimate 100,000 contracts, Elevate found that only 10,445 were “qualified contract documents,” while others were “not applicable” to the project, Van Nice said.

The next challenge was to extract relevant information, such as contact data or clauses that related to certain risks, from each contract. Using machine learning system could automate this identification— but first the technology had to be trained.

Patel recalled that the first time the team ran a batch of contracts through Kira’s AI engine, it could not identify or extract nine out of the 10 data points they needed. But with the next batch of contracts, the solution “found more information.” And after subsequent iterations, which helped to better train the AI, “we found it 98 percent accurate,” he said.

Training the platform, however, was not just a question of allowing Kira’s engine to continuously review batch after batch of contracts. The technology had to be told what was right and what was wrong in its identification of key data points.

To that end, Elevate deployed a team of lawyers who were familiar with the types and foreign languages of the agreements Baxter had to review. These attorneys continuously validated or corrected “the system as it was extracting the information,” Patel said.

Another speaker at the session—Sowmyan Ranganathan, senior director of legal operations at AbbVie—noted that this rigorous and extensive process was one of the reasons it’s difficult for many companies to train machine learning systems AI on their own. “They’re not going to have those kinds of resources that are needed,” he said.

To be sure, training AI with contracts is much easier when an organization already knows what its standard contracts encompass and which contracts it holds deviate from the norm. With that information at hand, its AI system can be more easily trained to understand the baseline of which contracts to work off.

But uncovering deviations within a contract database from the start is an onerous task. Patel noted that a company “can force [their] lawyers and business stakeholders” to edit contracts within a contract management platform. This way, the platform can track and provide metrics on the amount and type of changes being made over a large body of contracts. Yet such a process largely depends on many different in-house teams, whose adherence to the process may not be reliable as needed.

Far simpler, Patel said, was to review a contract for deviations when it first enters the organization. “What we typically see is at the point of storage, you have a contract team that says we’re going to load a contract and check for deviations,” he said.

Yet even while Baxter did not do that step ahead of time, the process using AI, even though it required extensive training, was still more cost effective than other options.

Van Nice, for example, noted in the end, the contract review project was around $500,000 dollars less expensive than what “other LPO providers that were going to do it manually [were offering],” and the AI saved an estimated 5,000 hours of work in total.

Copyright Legaltech News. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.