By Ken Adams
In some fields, the robots-running-rampant connotations of the term “artificial intelligence” might be somewhat justified. But when it comes to what LegalSifter offers—AI review of the other side’s draft contract—a dose of reality is in order.
The human component begins with deciding what the software looks for. At LegalSifter, we decide that in consultation with users and our law-firm “combined intelligence” partners. But if I think something matters, generally it goes on the list of “sifters” to be developed. (A sifter is a piece of software that looks for a specific issue.)
But the fun really starts with designing the specifications for each sifter. That involves figuring out how the issue targeted in a sifter is expressed in contracts. The specifications for a sifter might use one basic way, or “pattern,” to look for an issue, but a pattern might consist of several components, with each component potentially being expressed in various ways. For example, if one component of a pattern is the concept of obligation, anyone familiar with my categories-of-contract-language framework would know that obligations can be expressed in many good and not-so-good ways. We’d have to look for all variants that appear with any frequency.
Furthermore, a given issue might feature two or three patterns. For example, one sifter looks for statements that intellectual property isn’t being transferred under a contract. That sifter features a pattern that looks for language to that effect, but another pattern looks for provisions saying that one or more parties “reserve all rights” in intellectual property.
Once draft specifications for a sifter have been prepared, squads of lawyers pore over thousands of contracts to decide whether the specifications need to be refined.
As for the help text a sifter offers when the issue in question appears or doesn’t appear in a contract, I’ve been able to do a lot of it myself, as we’re still working on fundamental topics. I use my usual straightforward voice, with the aim of providing advice you can use. The more specialized topics—tech, intellectual property, real estate—I leave for other advisors.
The Role of Expertise
Our sifter-development process leads me to conclude that although LegalSifter sensibly enough plays the AI card—it’s the buzzword of the day—and although everything we do is enabled by technology, what’s front and center is expertise—my expertise and the expertise of everyone else who has contributed their contracts savvy to refining the product. The medium is technology, but the message is expertise.
LegalSifter has been transparent about who provides their contracts expertise. Plenty of other companies are sniffing around this space. Who provides their contracts expertise? I don’t know.
That poses two problems. First, whoever is advising them might not in fact have the necessary expertise. After all, it’s not as if there are a lot of contract-language specialists out there.
And second, even if they have suitable experts toiling away behind the scenes, that’s different from being seen to have the necessary expertise. Relying on someone else to tell you how to handle contract language requires a leap of faith. My years of writing, speaking, and teaching have allowed me to refine my work, but they’ve also kept me in the public eye, in a low-key way. With me, people know what they’re getting. I hope that makes it easier for them to make the leap to LegalSifter.
But it’s early days yet. I’ve been helping LegalSifter all of four months—my work has only just started.