As more enterprises automate away the tedium faced by in-house departments, a question looms: why haven’t in-house legal departments caught up? Internal legal processes for drafting, analyzing, and managing simple legal documents are still manual and tedious. What is stopping legal departments from automating away the pain?
As it turns out, a major barrier for adoption lies in the most common means of automation itself: Machine learning.
Contract Lifecycle Management (CLM) software streamlines and automates several stages of the contract lifecycle - from the initial drafting stages all the way up to negotiation, signature, and the final expiration of a contract. Most CLM platforms rely solely on one type of artificial intelligence (AI) for automation: Machine learning. Machine learning (ML) is a self-learning branch of AI, built for recognizing patterns in large pools of data. The problem with ML in the context of CLM software is that legal contracts are much more structured and nuanced than just any pool of data.
Although there are areas of the contract lifecycle where ML is useful – it can scan for key dates in third party paper or learn your negotiation habits – it is not a suitable tool for the more nuanced stages of the lifecycle. Mainly, contract drafting and analysis are too complex for a ML algorithm to perform reliably.
Both document drafting and analysis are stages of the lifecycle that require an artificial intelligence system to understand the content of a contract and the law itself. However, ML AIs are not trained to do this – they are simply built to recognize and act upon patterns. Thus, the way a ML AI would learn to write and read contracts is by analyzing thousands of documents. However, no matter how smart the AI, it is not possible to learn the entirety of the law just by reading legal contracts. As a result, ML AI is quite unreliable at writing or analyzing legal documents.
Though almost all industries rely on ML in some capacity for automation, its wide margin of error is especially problematic for legal work. Unlike a sales department, which would suffer no consequences for chasing a bad lead, the stakes of the legal department are far too high for Swiss-cheese contracts. Thus, while Sales is happy that their ML AI sends 70% of emails to the right people, Legal would explode if only 70% of contractual clauses were valid. Legal must therefore take extra precaution in their automation processes and use more tightly controlled AI for document generation and analysis.
In areas where ML fails, there are other types of AI to fill in the gaps. Instead of deploying self-learning AI such as ML for legal contract creation and analysis, enterprises should consider AI methods that are human-trained and tightly controlled. This way, instead of having the system learn the law by itself, legal professionals train the AI and set guardrails. A “Knowledge Graph” is a perfect example of a tightly controlled and human-trained AI.
“A Knowledge Graph is essentially a giant mesh of relationships linking entities together to represent real world facts, concepts, and ideas. By mapping the relationships that each clause has to each law, jurisdiction, and party, we weave a deep understanding of a contract into the system,” says Chetan Desh, Chief Technology Officer of CLM platform Advocat AI. “By using a Knowledge Graph AI instead of ML AI, legal insight is trained into the system, acting as a growing knowledge base over time. By understanding the content and context of a legal document, a Knowledge Graph AI can navigate nuance, allowing it to create solid legal documents and notify users when a contract needs attention.”
Though ML is the common favorite in the AI toolkit, the stakes of a legal contract are too high and the nature of the law too nuanced to leave this to ML alone. Thus, in areas of the contract lifecycle that are too complex and important for ML, legal departments should reach for the more suitable tool to get the job done. By deploying more controlled and trainable AI such as Knowledge Graph technology, legal departments too can automate away tedious tasks and start to focus on strategy.