Generative AI models are not new, but when ChatGPT was launched on 30th Nov '22, it felt like a seminal moment in adoption of AI. Since then, our teams have been playing around with the OpenAI chat interface / playground / APIs while our leadership has been busy talking to as many contract professionals as possible.
We had two objectives:
In this article, we start by sharing two examples of powerful application of ChatGPT on complex contract analysis work. We then share our ideas on the most impactful applications of Generative AI in contract management. Finally, we share a sneak peek of how we are integrating OpenAI APIs into our Microsoft Word based product.
Email us at [email protected] for access to private beta of Generative AI capabilities within Word
Ok, now lets talk about Generative AI and Contracts.
While a tool to fill up our screens with relevant clause text is interesting, a 'blank screen' has never been a problem for practitioners. There are umpteen templates, free and paid libraries of contract standards which professionals use to start a contract drafting exercise.
We believe a high value application for LLM in contracts is to analyze a individual contract clauses rigorously, converting them into its key components and suggest negotiation paths or alternate language texts
Here is a small example to illustrate that. We take a Share Purchase Agreement and ask ChatGPT to rigorously analyze its Termination clause:
Clause text from contract:
In the event that the Closing shall not have occurred with respect to a Buyer within five (5) days of the date hereof, then such Buyer shall have the right to terminate its obligations under this Agreement with respect to itself at any time on or after the close of business on such date without liability of such Buyer to any other party; provided, however, (i) the right to terminate this Agreement under this Section 8 shall not be available to such Buyer if the failure of the transactions contemplated by this Agreement to have been consummated by such date is the result of such Buyer’s breach of this Agreement and (ii) the abandonment of the sale and purchase of the Notes and the Warrants shall be applicable only to such Buyer providing such written notice, provided further that no such termination shall affect any obligation of the Company under this Agreement to reimburse such Buyer for the expenses described in Section 4(g) above. Nothing contained in this Section 8 shall be deemed to release any party from any liability for any breach by such party of the terms and provisions of this Agreement or the other Transaction Documents or to impair the right of any party to compel specific performance by any other party of its obligations under this Agreement or the other Transaction Documents.
Here is our prompt for ChatGPT:
"Assume you are an experienced attorney, US based and working on a buy side M&A deal in Pharma industry. Analyze the following termination contract clause by breaking it up into its key components. Determine whether it is permissive or mandatory. Further, highlight any restrictions it places on the buyer. Also, suggest negotiation paths and alternative language for the buyer.
Clause text: .....
...............
..................."
The results where pretty impressive! Something that would easily take more than a 20 minutes for an expert
Output from ChatGPT:
Analysis
Suggestions for Negotiations:
Alternative Termination Language:
The buyer may propose alternative termination language that is more favorable to their interests. For example, a "Material Adverse Effect" clause that allows the buyer to terminate if there is a material adverse effect on the target company.
Most attorneys (law firm or in-house) need to ensure that language in contracts that they are drafting or reviewing adhere to their organization standards. However, anyone who has done this work will tell you that comparing two, dense pieces of legal text for semantic similarity is anything but efficient.
Determining language similarity or drift from org standards or client's guidelines is another powerful application of LLMs in contracts.
Lets see how ChatGPT performs in this task (evaluating an indemnification clause in an investment agreement, in comparison to internal guidelines):
Clause text from contract:
The Sub-Adviser shall indemnify and hold harmless the Adviser from and against any and all claims, losses, liabilities or damages (including reasonable attorney’s fees and other related expenses) howsoever arising from or in connection with the performance of the Sub-Adviser’s obligations under this Agreement; provided, however, that the Sub-Adviser’s obligation under this Paragraph 5 shall be reduced to the extent that the claim against, or the loss, liability or damage experienced by the Adviser, is caused by or is otherwise directly related to the Adviser’s own willful misfeasance, bad faith or negligence, or to the reckless disregard of its duties under this Agreement. The Adviser shall indemnify and hold harmless the Sub-Adviser from and against any and all claims, losses, liabilities or damages (including reasonable attorney’s fees and other related expenses) howsoever arising from or in connection with the performance of the Adviser’s obligations under this Agreement; provided, however, that the Adviser’s obligation under this Paragraph 5 shall be reduced to the extent that the claim against, or the loss, liability or damage experienced by the Sub-Adviser, is caused by or is otherwise directly related to the Sub-Adviser’s own willful misfeasance, bad faith or negligence, or to the reckless disregard of its duties under this Agreement.
Illustrative guideline for Indemnification:
If the agreement unilaterally provides for the client to pay the legal expenses of the Disclosing Party if the client is the non-prevailing party, or even if the client is the prevailing party, in a litigation, then replace the language with a bilateral agreement for the non-prevailing party to pay the prevailing party’s legal expenses.
Here is our prompt for ChatGPT:
"Assume you are an experienced, US based attorney working on an Investment Sub-Advisory Agreement. One of the key guidelines to be followed while reviewing Indemnification and legal costs clauses in contracts is as follows:
If the agreement unilaterally provides for the client to pay the legal expenses of the Disclosing Party if the client is the non-prevailing party, or even if the client is the prevailing party, in a litigation, then replace the language with a bilateral agreement for the non-prevailing party to pay the prevailing party’s legal expenses.
Keeping these guidelines in mind, review the following clause and recommend any changes.
Clause Text:
......
......................"
The results were again pretty good for an expert to start with!
Output from ChatGPT:
The current clause provides for each party to indemnify and hold harmless the other party from and against any claims, losses, liabilities, or damages (including reasonable attorney's fees and other related expenses) arising from or in connection with their respective performance of obligations under the agreement. However, this obligation is reduced to the extent the claim is caused by or directly related to the other party's own willful misfeasance, bad faith, negligence, or reckless disregard of their duties under the agreement.
This clause does not follow the guideline to replace the language with a bilateral agreement for the non-prevailing party to pay the prevailing party's legal expenses. To align with the guideline, the following changes could be made:
For example:
"The Sub-Adviser shall indemnify and hold harmless the Adviser from and against any and all claims, losses, liabilities or damages (excluding reasonable attorney’s fees and other related expenses) arising from or in connection with the Sub-Adviser's performance of its obligations under this Agreement, to the extent such losses, liabilities, or damages are not caused by the Adviser's own non-willful misfeasance, bad faith, negligence, or reckless disregard of its duties under this Agreement. The non-prevailing party shall reimburse the prevailing party for their reasonable legal expenses incurred in any litigation arising from or in connection with this Agreement. The Adviser shall indemnify and hold harmless the Sub-Adviser from and against any and all claims, losses, liabilities or damages (excluding reasonable attorney’s fees and other related expenses) arising from or in connection with the Adviser's performance of its obligations under this Agreement, to the extent such losses, liabilities, or damages are not caused by the Sub-Adviser's own non-willful misfeasance, bad faith, negligence, or reckless disregard of its duties under this Agreement."
At ContractKen, we believe that LLMs are integral to the future of contract drafting, review, negotiations and management.
High value use-cases that we are working on:
If you're excited about the potential of LLMs in Contracts space and would like to collaboratively develop on ideas, talk to us. We are experimenting daily with 100's of different prompts and discovering new things!
We believe, the holy grail of Large Language Models applications in contracts is to enable a true Q&A type of chatbot facility for everyone in your organization to be able to query your entire contract corpus in natural language
This requires your documents to be converted to custom embeddings (think mathematical representation of contracts) and stored in a database. A cliched example is say when a senior leader (say your CFO) wants to know about all the customer contracts worth > 1MM up for renewal in next 6 months, having unfavorable termination clauses, she can just enter in natural language as:
"Give me a list of all our customer contracts worth > 1MM and renewing in next 6 months, and having an unfavorable termination clause"
and the system will generate a list of contracts, along with a list of basic metadata around those. Achieving that state will truly turn contracts into business assets and not just paperwork.
One of the challenges (and oft cited criticism of these models) is that they have been trained on contract text & concepts from almost entire internet (think EDGAR, books, articles, research papers, university lectures material, Ken Adams blogs, umpteen Twitter threads on contracts, etc.). As a result, the responses to tasks are sometimes unexpected or even less than accurate. Therefore, one needs to be thoughtful around using the output of these models i.e. as an assistant for your highly valued employees, not as their replacements.
Check out all our posts in this series: https://www.contractken.com/blog-category-generative-ai-for-contracts
More Like This
Day 6 of 20 - Contract Summaries Part II
Most comprehensive write-up on Contract Summaries - what are these, how to create them, how to use AI to generate different types of contract summaries and what are the key benefits and considerations in using contract summaries .
Read MoreDay 5 of 20: What is an AI Copilot?
We discuss why Copilot is the right paradigm to infuse AI into knowledge work. We delve deeper into factors driving the development and adoption of Copilots in all knowledge work areas like coding, analytics, contract review and drafting, copywriting, etc.
Read MoreCopy Paste Contract Clauses
Explore contract law's impact on partnerships, risks tied to clauses, and real-life cases. Leverage ContractKen's robust clause library for streamlined contract management, risk mitigation, and successful partnerships.
Read More