Using AI Services for Business Contracts: Is It Still There?

These AI products generally include a summary of the contract, a review of the contract for certain types of terms (such as limitations of liability), and any policies you have loaded into the system (e.g., the company is a vendor and has paid its maximum liability). We want to limit it to the total amount).

Technically, AI for contracts has its strengths and weaknesses, and while it is promising in some situations, it is unlikely to be useless (or even counterproductive) in other situations. Here are some predictions about usefulness, based on conversations with AI tech experts.

Contract summarization: Of the three things a company may ask of an AI contract product (contract summarization, critical terms review, and language revision), AI is weakest in summarization.

Technically, to perform the summarization, the AI ​​reviews the contract for the most important words, removes the irrelevant parts, and generates the summary. By doing so, you can make serious mistakes such as not making proper connections when encountering indefinite pronouns and other indefinite words. For example, using “this” by itself to refer to something else in the contract. Also, because of their placement, important usages of negative terms (such as “not”) may be omitted. Key quantitative values ​​(money, days to do something, etc.) can be misreported because aggregation omits important details and modifiers.

Therefore, it is risky to rely on AI-generated summaries to accurately report key aspects of contracts.

Contract revisions: AI should do a good job of identifying key contract terms and suggesting revisions to them to comply with contract policies entered into the system. Various tools, including a computer thesaurus, should be used to flag the types of terms you want to find or replace with languages ​​containing preset policies that rarely fail.

But can you trust your computer not to miss or mishandle critical issues? Lawyers can quickly verify replacement language by reviewing redlining changes in documents. But to see if the AI ​​has flagged all the key terms and conditions, lawyers have to go through the entire document, cutting savings. Savings therefore depend on risk tolerance.

When to use AI: AI contracts are best suited when the types of contracts being processed are common. In order for the AI-absorbed training data to provide good guidance, the contract types must be general.

The AI ​​can struggle when contract situations are rare, such as contracts with custom terms that are not often found in contracts. Flagging important terms can go wrong in rare situations because you don’t know what’s important. It may also fall short in suggesting revised wording.

In addition, AI could serve contract attorneys as brainstorming companions. For example, if a lawyer asked her AI to draft a certain type of contract, the AI ​​might suggest terms the lawyer never thought of or remembered to include.

AI won’t help with custom business terms: AI will provide nothing, or perhaps negative value, for custom business terms. Contracts are more than boilerplate, with cancellation rights, limitations of liability, and confidentiality obligations. Contracts typically include business terms such as prices, timelines, and descriptions of deliverables and services.

Unfortunately, these business terms are often vague or incomplete because they were written by business managers, not lawyers. AI has no real help in identifying or correcting these deficiencies.

Account for biases: If the training data used by the AI ​​consists of a corpus of contracts skewed to one side of the trade, the AI ​​should suggest modifications that are appropriate for the benefit of the other side of the trade. may not work.

For example, if you are representing a borrower in negotiating a commercial loan agreement, if the AI ​​learns about a commercial loan agreement drafted by the lender, it will flag critical terms or suggest appropriate amendments to protect the borrower. or may not work. Specific contract policies entered into AI should be handled appropriately. However, due to training biases, the AI ​​can miss issues not mentioned in the contract policy.