Why Does IntelAgree Use Machine Learning?
IntelAgree leverages machine learning models to help lawyers do impactful work, not busy work.
Since most agreements follow similar structures and contain certain terms, our models are designed to help clients access crucial information — like renewal dates, payment terms, and termination clauses — faster.
These models are meant to be an aid, not a replacement — and in order to make your machine learning model successful, you’ll need the right training, guidance, and expectations. In the next two sections, we explain what to anticipate during and after implementation.
Did You Know?
Companies that use intelligent CLM platforms reduce delivery turnaround time by 40% per contract.²
By 2024, companies using advanced contract analytics solutions will reduce manual contract review by 50%.³
Compared to traditional models, companies see a 60% cost reduction with intelligent CLM implementation.⁴
Machine Learning: What to Expect During Implementation
The Four Machine Learning Outcomes:
After you have trained a model for an attribute, you can use the model to find that attribute in a new contract. The way a model “reads” a contract is by breaking it up into individual tokens. In contracts, a token is any set of characters that belong together; this could be a word, abbreviation, or even punctuation. When a contract is imported into IntelAgree, the model sees the contract as a collection of tokens. As each token passes through the model, it is assigned one of four outcomes:
True Positive - Success! The attribute you trained your model for recognized the language through pattern recognition and marked the attribute.
True Negative - The attribute did not exist in the document and the model correctly did not mark any text.
False Positive - Machine learning recognized language that was similar to the language from your trained attribute, but it was the wrong text.
False Negative - Machine learning did not capture an attribute in a document that it existed in.
To rectify false positives and false negatives, undo the machine learning’s markup and manually correct the attribute. Every time the attribute is marked correctly, use each of the correct data points to retrain the model. This is critical in creating well-trained models to directly identify tokens. This is also important regarding new data; if your new data is set up differently than your previous data, such as a third-party paper, additional retraining must be done to attain the desired result.
See IntelAgree In Action
Still have questions about how IntelAgree uses machine learning to accelerate contract management? Schedule a demo to see how IntelAgree makes easy work of managing contracts.
Download the PDF
Sources
1 Driving Better Fiscal Management and Revenue Recognition with Enterprise Contract Management Technologies.” World Commerce & Contracting (formerly IACCM), 15 Dec. 2008, www.worldcc.com/Resources/Content-Hub/View/ArticleId/7326/Driving-better-fiscal-management-and-revenue-recognition-with-enterprise-contract-management-technologies
2 Gartner, Inc. “Our Top Data and Analytics Predicts for 2021.” Andrew White, 12 Jan. 2021, blogs.gartner.com/andrew_white/2021/01/12/our-top-data-and-analytics-predicts-for-2021/.
3 Lorenzo, Lori. Tech Bytes Part 4: Intelligent Contract Life Cycle Management. Deloitte, www2.deloitte.com/content/dam/Deloitte/us/Documents/about-deloitte/us-about-deloitte-tech-bytes-part-4.pdf.
4 Lorenzo, Lori. Tech Bytes Part 4: Intelligent Contract Life Cycle Management. Deloitte, www2.deloitte.com/content/dam/Deloitte/us/Documents/about-deloitte/us-about-deloitte-tech-bytes-part-4.pdf.