AI and centuries old contract management
Some of the most exciting opportunities for LLMs involve optimizing the most onerous parts of work. I’m referring to the tedious, nuts-and-bolts tasks that just about every company deals with on a daily basis to run their business. A lot of this work is time-consuming and finicky and still requires human expertise to parse and review.
One such area is contract negotiation and management, a complex and highly necessary part of business that every company on the planet deals with. After all, every dollar that goes in and out of a company is represented by some sort of contract.
One company that’s using LLMs to ease the process of contract negotiations is Icertis, a startup founded in 2009 that counts many Fortune 500 companies as their clients. My guest this week is Icertis’s co-founder and CTO Monish Darda who, in addition to providing me with some valuable insight into the world of contract negotiations, shared how Icertis is using LLM technology to make the negotiation process easier.
The German trucking company and the one-hundred-year-old contract
To understand how LLMs could benefit the process of contract management, it’s important to understand how painstaking the existing system is for many companies, including one of Icertis’s very first clients, the Mercedes-Benz trucking company Daimler.
Founded in 1926, Daimler does a lot of business with manufacturers whom the company has worked with for almost a hundred years–which means that, when Icertis first began working with the company, many of Daimler’s contracts were based on one-hundred-year-old agreements. “They were trying to figure out the contracting process on the sourcing side, especially around buying material,” said Monish. “They spend around $80 billion every year on material, and many of their contracts have evolved over decades with thousands of amendments.”
Daimler was accustomed to its contracts lasting a very long time, but there was a catch: “They often found that once these contracts were put into practice, the original intent represented by the contract was very different from the reality of the on-the-ground transactions,” said Monish.
This was when Monish and his co-founder, Icertis CEO Samir Bodas, realized they could leverage AI and machine learning to create contracts that better represented the interests of companies based on pre-existing agreements and negotiations.
Using LLMs to generate and drive contracts
But this proved to be far from straightforward. The nuances of contract law can be particularly tricky for machine learning algorithms and LLMs to parse. Even a seemingly easy task, such as identifying the expiration date of a given contract can be difficult, since contracts often include terms that go into effect only once certain requirements are met.
“Our objective when we started using AI was to understand the interaction of the contract on external systems and with external people. If you get a contract, you should be able to figure out what other contracts it’s related to, what type of contract it is, and what kind of attributes it has based on agreements you’ve already negotiated with other clients,” said Monish. “Our first AI offering was to discover attributes in a contract.”
This AI offering has evolved into a critical part of Icertis’s contract intelligence platform that helps companies automatically assemble and analyze contracts based on pre-existing agreements and clauses. With the advances in AI, Icertis has recently introduced LLMs into its offering along with a generative AI application for enterprise contract management called Contract Intelligence Copilots. Copilots uses third-party LLMs running on Microsoft Azure OpenAI, structured data from the enterprise, an enterprise data lake, and a vector database to generate contracts. One of Icertis’ biggest advantages is its huge repository of contract data to train its models on.
LLMs have made generating, analyzing, and monitoring contract clauses much easier for customers. “Often, when you use machine learning on simple things like words you get misled by spelling mistakes or complex legal terms,” Monish said. “Now you can go into an LLM that is already trained and get a better understanding that can flow into discovery.”
Implementing LLMs has worked well for Icertis’s team because it had a specific use case and existing tools to convert the outputs of this new technology into business outcomes. “This is where a lot of startups are missing the point in my mind,” Monish said. “When you use AI, and specifically when you use LLMs, it’s not enough to use the model and get a result. It’s also important to integrate the models into a structured workflow.”
What enterprise customers think about using LLMs to generate contracts
Often I hear that companies receive a lot of pushback over privacy concerns from enterprise customers once they introduce AI, especially LLMs, into their products. But Icertis received the exact opposite reaction when they showcased an early version of Copilots to its customers–which is a testament to how valuable customers believe their product could be.
“The excitement was extreme,” said Monish. “People expected magic.”
From the beginning, Icertis’s clients had an innate understanding that AI could process and generate contracts with more speed and accuracy than a multi-person legal team. “They seemed to catch onto the concept that the manual process generates a lot of mistakes,” Monish said. “They also realized that the only way to have a comprehensive understanding of the contracts they previously signed was to use machine intelligence.”
“Now, we have the ability to say, ‘What you’re negotiating here is very similar to this contract you agreed to five years ago,’” he continued. “Instead of spending hours negotiating these agreements from scratch, AI sets up a foundation from which you’re better equipped to negotiate terms that work.”
But even despite this initial enthusiasm, Icertis’s clients are not yet ready for contracts to be entirely auto-generated. After all, AI technology still needs humans in the loop–which is why human review is an extremely important part of the contracting process. Today, Icertis generates its contracts on a clause-by-clause basis. “People in enterprises are not yet comfortable with a contract that’s entirely auto-generated,” he said.
Sounds like there’s a lot of exciting innovation ahead in the field of contract negotiation.