On Day 2 of 20 in our series on 'Generative AI for Contracts', let see:

How Generative AI can be deployed to get analyze specific contract clauses deeply.

Since Covid19, one clause that has been under the scanner more than anything else is 'Force Majeure'.

In this example, we discuss a technique called as 'Chain of Thought' to illustrate how LLMs can perform reasoning tasks in contracts.

Lets take a look at a Supply Agreement between ABC Industries, Inc., a Florida corporation (“ABC” or “Customer”), and XYZO Extrusions, Inc. a Delaware corporation (“XYZO” or “Supplier”). The agreement shall govern various purchase orders that will be raised for supply of various products from supplier to the customer.

Following is the Force Majeure clause:

No party to this Agreement shall be liable for non-performance hereunder resulting from: severe weather conditions; war; riots; civil disorder; earthquakes; any law, order, proclamation, regulation, ordinance, demand or requirement of any governmental agency; or any other condition or occurrence whatsoever beyond the control of such party, but only to the extent performance hereunder is prevented by any such condition. If the performance of this Agreement is prevented by reason of any such event, (a) the party whose performance is prevented, shall give prompt written notice to the other party of the event and shall be excused from performance, but only to the extent prevented; provided, however, that the party whose performance is prevented shall take all steps to avoid or remove such causes of nonperformance and shall continue performance whenever and to the extent possible; and (b) if it appears that a time for delivery or performance scheduled pursuant to this Agreement will be delayed for more than six (6) months, the party receiving notice under subsection (a) above shall have the right to terminate, by written notice to the other party, any portion of this Agreement covering the prevented performance, and the obligations and liabilities of both parties with respect to such portion of the Agreement shall thereupon lapse and terminate, except to the extent such obligations or rights are intended to survive pursuant to this Agreement.

So, lets start by asking ChatGPT about various components of a Force Majeure clause:

‍

‍

We then take the output of first prompt, and use it to decompose the 'Force Majeure' clause in supply agreement as follows:

‍

Lets now bring our client's preferences related to Force Majeure and get the model to understand those and compare the contract clause being reviewed against those.

Here is a sample of client's preferences:

‍

Now, we'll further use the output of previous prompt and with some careful prompt language, as the model to ingest the above information, analyze the force majeure clause text against these standards and produce the final output in yet another neat table.

‍

‍

Finally, what we get is the output table as below:
‍

This is now a structured analysis of various components of the contract clause against your client's preferences.

ContractKen strongly believes that by using techniques like chaining prompts and few shot learning, LLMs can be extremely useful in faster and deeper analysis of contract clauses

We are building this kind of capability within Microsoft Word so that contract professionals can perform this level of analysis without leaving their preferred tool (Word, not ChatGPT ;) )

‍

We will cover some of the concepts used here in detail in upcoming posts.

‍

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

‍

Further readings:

• https://www.contractken.com/articles/7-most-negotiated-clauses