I have been visiting the same Starbucks, in Crystal Springs Shopping Center, for the past 15 years. At first, I would arrive early, on my way to work, ready to claim my cup of Pike Place Roast, an improvement (perhaps) over the coffee from my Keurig. On the patio stand a few tables and chairs, and, invariably, patrons, in their forties or older, sit, enjoy their coffee beverages, and discuss, if not pontificate on, the issues of the day. Thus, to my surprise, for the past few weeks, as a means of varying my pandemic work-from-home routine, at the midway point of a morning hike/walk, I have been enjoying lattes and cappuccinos (not quite as good as the ones from Trastevere, but functional indeed) with friends, opining on our local education system (yes, public schools can work, if sufficiently funded) and matters of a more global nature (if Chelsea only had a great attacking midfielder and playmaker), before my daily conference calls. Put otherwise, the future is here.
I have given some thought to the future of law practice. If you practice, as I have, for more than 20 years, it is hard not to do so. In 2018, I wrote a somewhat wonky article in New Matter (IP Section, California Lawyers Association) on Licensing in the Age of Artificial Intelligence. To be fair, part of it was grounded in the present. Another part, however, was more of a thought experiment than anything. I wrote, in part:
AI raises a second, thought-provoking issue for licensing lawyers, namely the potential automation of the license review and negotiation process. On one level, this has been happening for some time, in the form of document comparison software. Going one, or a few, steps further, once a company has established its policies, AI software could grade a counterparty’s markup of the company’s license agreement template. Perhaps, if properly trained, AI could, with a relatively high degree of confidence, determine if the language in the markup falls within the bounds of the accepted fallback positions developed by the template owner. Some human (attorney) supervision would be required, but, again, this could be workable. Imagine, however, taking this one step further, by giving the software program agency, permitting it to negotiate, based on pre-established rules (e.g., no reverse indemnities in a purchase agreement, where the principal is the purchaser or two enumerated jurisdictions permitted for choice of law, with the requirement that venue match governing law). At the program’s immediate disposal would be a repository of all U.S. state and federal laws as well as published cases. After the “chess match,” attorneys for the licensor and licensee would review the resulting agreement, discuss with their respective clients, fine tune if and as necessary, and close the transaction.
This would be interesting indeed. Like many attorneys, I have been in negotiations where the counterparty might as well have been an automaton, whether blindly citing (and not explaining the rationale for) a policy as a basis for not accommodating a proposed change (bearing in mind, of course, that many policies have exceptions) or reflexively insisting on mutuality (where the risks to each party are different, qualitatively, if not also quantitatively), or focusing exclusively on spends rather than also giving due consideration to inherent risk as presented by the transaction. With this in mind, I might even volunteer to be a beta user for the software program described in the paragraph above.”
Imagine, now, that this simulated negotiation yielded a more optimal solution for both parties. Imagine, further, that at some point humans could be disintermediated from the process, saving time and money. Would you sign up for such a process?
Let’s assume, initially (before disintermediation), this process could make me twice as efficient, halving my typical workday from 10 hours to 5. As an in-house attorney, I am quite sure than I could find another 5 hours of work (or, rather, the work would find me), given my backlog. Perhaps, although I doubt it, my day would be spent coding inputs and validating outputs. Thus, the transactional lawyer of the future would become the pilot of today: laser-focused on take-off and landing and otherwise, but for some troubleshooting and adjustment, if a deal encountered unexpected turbulence, letting the plane (via sophisticated software) fly itself. If I were still in private practice, and not billing on a fixed fee basis, I would need to develop new, value-added services, to compensate for the time I could no longer bill. Now, as a client, this kind of automation could be salutary (in some respects), although I could imagine at least a few, if not all, State Bars declaring this the unauthorized practice of law, barring substantial attorney supervision. More concretely, ABA Model Rule 5.3 (Responsibilities Regarding Nonlawyer Assistance) would require a substantial overhaul. Regardless, AI in a law practice is probably a matter of when, not if.
In my above-referenced article, I do note that more complex transactions, like collaboration or joint development agreements, would be a more challenging substrate for this experiment. But, then again, I could probably devise a questionnaire and concomitant rules (of negotiation) there as well. For example, if (1) the agreement has more than three parties, (2) I do not have any background IP in the planned collaboration space, and (3) my principal goal is to ensure freedom to operate, I could hard code a preference for joint ownership of any jointly developed IP. Surely, it would take time to hard code these preferences, but this is not an impossible task.
My recent research on this topic has brought me to another somewhat wonky topic: Rules as Code. Simply put, in addition to representing laws in legal language, laws could be coded as well (in Python, etc.), allowing users to, arguably, better query and understand them, and to better model potential impacts, than the current (18th century) model. A recent presentation from the Australian Society for Computers and the Law (AUSCL) proved quite illuminating: https://www.youtube.com/watch?v=H3YgOh4WU7U&feature=youtu.be
I have started exploring this concept, including documentation on Open Fisca, a platform for transforming laws into code: https://openfisca.org/doc/coding-the-legislation/10_basic_example.html
I am beginning to regret (slightly) dropping my one CS course (Fortran), as a freshman in college, but in the spirit of continuous learning, it is never too late to pick up a new skill.
For prescriptive rules, it is possible, I suppose, that rules as code could work just as well as (if perhaps not better than) our current paradigm. For principle-based rules, i.e., those grounded in judgement, as noted quite astutely during the above-referenced AUSCL presentation, there is no reason why the rules engine could not offramp to a skilled practitioner, to apply appropriate judgment. Whether coded rules would be fairer is open for debate. Again, conceptually, this is all very interesting, and a key next step for me (besides learning Python) is to give more thought to the desired end state of these efforts (before working backwards).
Lastly, I am exploring scholarship on whether or not any one modal representation of a law is better than another. More to the point, laws have been expressed in traditional written form because to date (arguably) this has been the best way to record and communicate laws. This is not to the say that the essence of the law cannot be expressed in another form, such as computer code, just as well, if not better. The following article (The Law’s New Language?) explores this concept:
So, where does this leave us?
For starters, law schools would be well-advised to offer a coding course, in addition, of course, to a course in technology transactions. Specialization and commoditization of (some) legal services are longstanding trends, so identifying new value-added services will be key. Fragmentation in the delivery of legal services will continue, with AI as an adjutant to new service models, although the changes will take longer than most people think. Lastly, Rules as Code will not be a passing phenomenon, although so long as it is driven by software engineers, and not embraced to some extent by legal practitioners, its future might be limited.
Comments welcomed.
The Peninsula Hiking Club 