On Day 3 of 20 in our series on 'Generative AI for Contracts', lets talk about Contract Summaries.

How Generative AI can be deployed to build accurate Contract Summaries efficiently.

When it comes to Contract Summaries, there are two types-

• A legal summary of a contract typically provides a detailed analysis of the legal terms and provisions of the contract. It is written in a highly technical and legalistic language, and is primarily intended for lawyers and other legal professionals who need to understand the legal implications of the contract. The legal summary may include detailed discussions of the governing law, jurisdiction, indemnification, warranties, and other legal terms that are critical to the agreement.
• On the other hand, a commercial summary of a contract is written in a more user-friendly language that is accessible to non-legal professionals, such as business executives or stakeholders. The commercial summary focuses on the commercial terms of the agreement, such as the scope of services, pricing, payment terms, and deliverables. It may also include a summary of key business risks, obligations, and benefits, and is designed to help the business stakeholders to understand the commercial implications of the agreement.

Typically, all organizations have developed some sort of template or checklist that the summaries follow. Usually, it contains a combination of important legal and commercial considerations. And its usually 1-2 pages (so that busy executives can approve the contract on its basis). Using Generative AI to assist in population of such a summary template is a very powerful usecase. We will look briefly at how this is done, later in the article.

Before we delve into the details of how to develop these customized contract summaries, lets discuss what a summary is.

When it comes to generating summaries using generative AI models, there are two main approaches: abstractive and extractive.

Abstractive Summaries:

Abstractive summarization involves generating a summary that uses natural language to express the essence of the original text. This approach involves analyzing the text and generating a summary that captures the main ideas and themes of the original text, while also using new words and phrasing to create a more readable and engaging summary. Abstractive summarization is more challenging than extractive summarization because it requires the AI model to understand the meaning of the text and to be able to generate new language that accurately reflects that meaning.

‍

Extractive Summaries:

Extractive summarization, on the other hand, involves generating a summary that uses exact words and phrases from the original text. This approach involves identifying the most important sentences or phrases in the original text and using them to create a condensed summary. Extractive summarization is easier to implement than abstractive summarization because it does not require the AI model to generate new language. However, extractive summarization may not capture the full meaning of the text and may result in a summary that is less engaging and readable than an abstractive summary.

‍

Lets take a look at how ContractKen's Word Copilot helps you create 3 types of summaries right inside Word - Abstract Summary, Key Terms Summary and Key Issues Summary.

Typically, most people have their specific view of how a summary should look like. To that end, we work with our customers and create 'custom summaries' with AI. The process for that roughly looks like this -

First, you'll need a (largely) fixed template or checklist that you're looking to populate for new contract documents. Then you'll need to understand all the elements in the summary in detail (by reading through the relevant portion of the contract and with the help of an expert). Its important to understand the nuances of language which represent specific concepts / terms in the summary document. Write specific questions (as you'd ask your legal assistant) around those specific topics. Convert those into prompts you will use with an LLM. However, how does the LLM know where to search for the specific answer (trying to scan through the entire document each time for a specific question is not a very efficient algorithm). Hence, its better to use mathematical representations of the text (also known as 'vector embeddings'). This step allows your specific questions / terms and concepts to be searched efficiently across the document. LLM then picks up the relevant text (typically ranked highest in similarity to the concept being asked for) and answers with a precise sentence. Finally, you need some plumbing to assemble the answer sentences in a particular format to give you an output which you are used to reading.

Reach out to us at hello@contractken to discuss how you can leverage Generative AI to create custom summaries for your contracts.

‍

Check out all our posts in this series: https://www.contractken.com/blog-category-generative-ai-for-contracts