The Blue-Sky Bolt and the Role of Insight in the Legal Realm
Recently someone pointed out a typo in the fifth edition of A Manual of Style for Contract Drafting. (In a 667-page book with lots of technical stuff, there will be typos—I can cope!) In an email to them, I joked about awarding them the “Crimson Lightning Bolt for Glitch-Finding Valor.” That got me wondering what such a medal might look like. Because I now have in my corner Moh Soumeur, designer of my monkey illustration (see this blog post), I was in a position to indulge in this sort of fantasy. I changed crimson to sky blue, to match the cover of the fifth edition of MSCD, and I suggested to Moh some other design elements. Here’s the result:

Unlike the thunder cloud and rain that come with lightning, the Blue-Sky Bolt brings you the blue sky, dispelling the murk in a flash. I like it enough that I might use it to acknowledge any helpful insight, suggestions, or questions that come my way. I’ve already ordered enamel pins.
But thinking about the Blue-Sky Bolt got me thinking more generally. When it comes to contract language, most of us don’t have a lot of time for human insight. We might be copy-and-paste monkeys (see this blog post for my definition). We might be marketing types, satisfied with appearances. We might be social-media influencers and cheerleaders (see this blog post), perhaps more invested in the conversation than in the outcome. Or we might be in thrall to inertia (see these blog posts).
And now, looming above all else, we have the large-language-model (LLM) version of artificial intelligence, currently exemplified by GPT-4. (See this blog post by Casey Flaherty for some context.) It appears that GPT-4 will further narrow our scope for deploying human insight in the legal realm.
Take legal research memos. I reckoned myself a wiz at legal research memos, but they tend to be wasteful, with armies of junior lawyers and summer associates tackling the same topics over and over. Even decades before GPT-4, I sensed the commodity nature of legal research memos: when I wrote them, I felt as if I were on a production line, applying the same reasoning tools, whatever the topic. So even though the unaided human calls on insight to write legal research memos, I wouldn’t be sorry to have GPT-4 elbow us out of the way to some extent.
So where’s a lawyer to deploy insight? You can already find plenty of opinions on this, but presumably the center of gravity will increasingly shift from the lawyer version of manual labor to helping the client make decisions.
This could have been bad news for me, as I live for small flashes of insight in my research and writing. My pragmatic version of scholarship—slowly moving from confusion to being able to offer an alternative to the conventional wisdom—is about as exciting as things get for me, but it’s labor-intensive. By contrast, I was never particularly interested in, or much good at, the expediency-driven process of holding the client’s hand as they walk from Point A to Point B.
But happily for me, the stew of dysfunction that is traditional contract language is largely immune to having GPT-4 or any other LLM be the source of truth on contract language. LLMs repurpose or emulate what’s out there, and there’s not much future in repurposing or emulating dysfunction. (For more about that, see this blog post.)
In the wake of Carrick Capital Partners’ investment in LegalSifter (go here for more about that), we’re contemplating our next steps in both contract review and contract creation. In contract review, there will certainly still be a role for artificial intelligence, as there has been all along. But in both spheres, whether we succeed will be a function of how well we harness the lightning bolt of human insight, in addition to whatever artificial-intelligence-driven repurposing and emulation we take advantage of.
(This is a revised version of a post on Ken Adams’s blog, at www.adamsdrafting.com.)
Recent Posts
See AllRecently I noticed this article on Artificial Lawyer. The title is Generative Legal AI + “The Last Human Mile”, and it’s about limits to applying AI to legal work. It says this: The last mile problem