Generative AI Embraces 'It Depends' -- Aw, It Thinks It's (Lawyer) People!
Maybe it's a stretch to consider lawyers as people, but you get the drift.
Unless you’re trying to commit malpractice or taking the bar exam — everyone knows the first answer to most every legal question is “it depends.” That’s a feature, not a bug. The whole point of an advocate is to find the nooks and crannies within the law that can benefit the client. If the law were so fixed as to invite one “answer,” we wouldn’t be sending graduates into six-figure debt and Clarence Thomas wouldn’t be getting so many free vacations.
When generative AI splashed onto the legal scene, it arrived with a lot of bold claims about quickly and efficiently providing lawyers “the answer” to everything from legal research to perfectly drafted contracts. And these large language models will certainly deliver an answer — right or wrong — unless someone takes the time to build guardrails to stop it. Because, as a wise person (I believe Christine Lemmer-Webber) once dubbed GenAI, it’s “mansplaining as a service.”
But beyond its gusto to give the user finality, early legal applications of the technology leaned into the rhetoric of “answers.”
What a difference a year makes.
One word that kept rising to the top of conversations at this year’s ILTACON was, “iterative.”
I recently employed the term in an article after chatting with the folks from Paxton AI. During that conversation I’d picked up on a subtle shift from earlier industry talk that positioned AI as “one-and-done” prompting to an acceptance of the chatbot interface as the basis of a, you know, chat. A back-and-forth that a senior might have with a junior where queries are refined or recast over the course of the dialogue.
Words convey mindsets and as this phrasing kept cropping up, it’s clear that AI’s development is being carried along the path of incorporating the tech into the give-and-take workflow. Everlaw founder and CEO AJ Shankar explained:
It’s an iterative process, and I think the kind of question-and-answer oracle “here’s the answer” stuff misses the point. In fact, it’s a very nuanced, subtle, iterative experience of getting from evidence to a compelling narrative. And that is kind of how we designed this. You get it to do a first draft, you can have it rewrite segments, you can just dialogue with it, you can take deposition information, integrate it in there, you kind of have to just take a while. It is meant to supplement, it’s not gonna replace… we’re much more interested in like, how can this just help you do your job better, be a foil for you.
Artificial intelligence as foil takes the dialogue approach even further than the partner-associate relationship toward a Devil’s Advocate in a box. Or maybe a Hegelian dialectic starter kit.
At a roundtable event hosted by Level Legal, bringing together media and practitioners to talk about the pressing issues in eDiscovery, “iterative” popped up yet again, as a firm discovery specialist flagged the concept as a baseline for the responsible use of GenAI. Attorneys cannot — ethically — say “Anthropic Take The Wheel,” though they also should not eschew tools with the potential to make marked efficiency (and potentially even substantive) improvements to the discovery process. Leaving the profession back at the concept of iteration: constantly pinging, assessing, revising, and reassessing the tool as it enters the legal workflow.
Again, not unlike lawyers would’ve done with review attorneys or QC specialists when ALSPs dominated the landscape.
The meaning of the law exists in a state of flux until an attorney achieves a high enough level of confidence that “it depends” transforms into “this is at least what it means for our client today.” To the extent GenAI has a role to play in that process it has to be designed — and used — with an eye toward incremental, evolutionary use.
Last year at ILTACON, that wasn’t the dominant discourse surrounding GenAI. This year it was.
The fruit of 12 months engaged in an iterative process.
Earlier: AI Feature Embraces Traditional Role Of Asking First-Years To Go Back And Do A Lot More Research
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news. Joe also serves as a Managing Director at RPN Executive Search.
