Because the public launch of ChatGPT late final yr, I’ve been inundated with solicitations from corporations providing new AI instruments for enterprise contracting. Are they prepared for prime time?
These AI merchandise typically declare they will do three issues: summarize a contract, evaluate a contract for sure sorts of phrases (e.g., limitation on legal responsibility), and recommend amendments to a contract based mostly on insurance policies you load into the system (e.g., your organization is the seller, and also you want to restrict its most legal responsibility to the full quantity paid).
Technologically, AI for contracting has strengths and weaknesses that make it promising for some conditions however unlikely to be useful (and even counterproductive) in others. Primarily based on speaking with individuals who have experience in AI know-how, listed below are some predictions on usefulness.
Contract summarization: Of the three issues an organization would possibly name on an AI contracting product to do (contract summarization, evaluate for vital phrases, and language revision), the AI will probably be weakest in summarizing.
Individuals are additionally studying…
Technologically, to carry out summarization, the AI will evaluate the contract to attempt to verify which phrases are most vital after which remove unimportant components to generate a abstract. In doing so, it would make vital errors, akin to not making correct connections when encountering indefinite pronouns or different indefinite phrases, akin to stand-alone use of “this” to check with one thing else within the contract. It may additionally omit a key use of a negation time period (akin to “not”) due to its placement. It would misreport a key quantitative worth (e.g., a financial quantity, the variety of days to do one thing) as a result of its summarization cuts out key particulars or qualifiers.
Thus, it’s dangerous to depend on an AI-generated abstract to report the vital elements of a contract precisely.
Contract revision: The AI ought to do nicely at figuring out vital contractual phrases and suggesting revisions to them to adjust to the contracting insurance policies you enter into the system. It ought to have the ability to use numerous instruments, together with laptop thesauruses, to hardly ever fail in flagging the varieties of phrases you wish to discover and in substituting language containing your preset insurance policies.
However are you prepared to belief the pc to not miss or mishandle essential points? An legal professional can test the alternative language shortly by wanting on the doc’s adjustments within the redlining. However checking to see if the AI flagged all of the vital contractual phrases would require an legal professional to evaluate the entire doc, which cuts down on financial savings. Thus, your financial savings will rely in your threat tolerance.
Finest conditions for AI utilization: AI contracting will work greatest when the kind of contracts processed are commonplace. The kind of contract have to be commonplace for the coaching information absorbed by the AI to offer sufficient steering to it.
If the contracting state of affairs is uncommon, akin to a contract containing customized phrases not regularly present in contracts, the AI might battle. It would do poorly at flagging vital phrases as a result of it doesn’t know what’s vital in unusual conditions. It additionally would possibly do a poor job of suggesting revised language.
Past that, the AI could be useful to a contracts lawyer as a brainstorming buddy. For instance, if the lawyer asks an AI to draft a contract of a specific sort, the AI would possibly recommend phrases the lawyer wouldn’t consider or keep in mind to incorporate.
AI gained’t assist with customized enterprise phrases: An AI seemingly will present no or maybe detrimental worth regarding customized enterprise phrases. A contract is not only boilerplate, akin to termination rights, limitation of legal responsibility, and confidentiality. A contract normally incorporates enterprise phrases, akin to costs, timelines, and descriptions of deliverables and companies.
Sadly, these enterprise phrases are sometimes obscure or incomplete as a result of they’re written by enterprise managers moderately than attorneys. The AI isn’t going to be materially useful in figuring out or fixing these inadequacies. When contractual relationships fail, it’s typically due to a scarcity of readability and element in enterprise phrases.
Take into account bias: If the coaching information utilized by the AI consists of a corpus of contracts slanted towards one facet of a deal, the AI would possibly do poorly in suggesting revisions acceptable to the pursuits of the opposite facet of the deal.
For instance, should you characterize a borrower in negotiating a business mortgage settlement, and if the AI discovered on business mortgage agreements drafted by lenders, it would do a poor job of flagging vital phrases and suggesting acceptable revisions that shield the borrower. It ought to nonetheless do a great job in dealing with particular contracting insurance policies you enter into the AI, akin to putting any probability of a confessed judgment. However, attributable to coaching bias, the AI would possibly miss points you didn’t put in your contracting insurance policies.
John B. Farmer is a lawyer with Main-Edge Legislation Group PLC, which focuses on mental property regulation. He could be reached at www.leadingedgelaw.com.