Urban Dictionary defines “automagical” as “something that happens automatically, but you can’t quite explain how it happens, or the explanation is too complex.” It’s a word that brings a bit of sparkle to my mind when I imagine it and evokes exactly that sentiment and more —an assumed underlayer of complexity, iridescence, desirability, and progress. In speaking with friends and colleagues and being part of the wider industry conversation these last few months around generative AI (gen AI), legal tech, and the debate and dialogue around “what’s next” for us lawyers, it seems that legal tech may be perceived exactly in that way by many of us. The automation is risky and scary. The explanation is tough to grasp and pin down. And the way in which it gives us answers is too risky to be meaningfully considered as a tool in our legal tool kit as expert practitioners.
I’d like to flip that perspective and posit that we should see it differently. The legal tech we’re starting to see from various industry providers holds tremendous promise. Not only in automating, accelerating, and enabling the work we do, but also in reshaping how we work, the amount of time spent on the lower-level versus higher-level work we do, and more broadly, the way in which we can reshape our delivery vehicles as lawyers and offer clients a better, sharper, more accurate work product that captures a wider scope of coverage and perspective.
In this thought piece, I’d like to suggest that, organizationally, in-house legal departments and law firms alike would greatly benefit from taking a thoughtful, intentional, and deliberate approach to adopting legal tech. I’d like to map out a journey showing that it can be easier than anticipated, with the yield from the exercise setting you up for success and providing dividends from a competitive, skills-focused, and client satisfaction perspective.
First thing’s first: Let’s form a strategy
Step one in identifying the legal tech an organization needs is to identify what the organization needs. A number of legal tech vendors and value propositions are evolving, with use cases and capabilities quickly developing. What I’ve noticed is that the order of operations in piloting and, ultimately, adopting legal tech, and organizationally setting up for success, is critically important.
Rather than being curious about the tech and going straight to the tech vendors to explore the art of the possible, the more effective strategy is to first identify the use cases that are broadly available in the industry (for example, what can gen AI-powered legal tech do across legal services and practice areas, specifically), and second, consider those use cases, as they may be applicable to your legal organization. Select the ones that are of high value and scalable across your law firm or department (three to four leading use cases are suggested). Third, with those leading use cases in mind, identify the leading tech vendors matching those offerings, and consider whether you have any internal capabilities in development that could be applied to match those offerings. This is where the build versus buy versus partner consideration becomes increasingly important—that is, do you build your own tech capabilities internally and fine-tune the large language models (LLMs) (meaning you customize a pretrained LLM to a specific task by training it on an identified, smaller data set), do you partner with data scientists or legal tech vendors to cobuild and codevelop tech, or do you buy (that is, license) the tech from existing tech vendors and tailor it to the needs of your department?
Alongside the rise of innovation, legal operations, and leading-edge knowledge management leads, a number of law firms are now creating dedicated AI and machine learning departments focused on the application of AI (which is, increasingly, simply becoming generative AI) to their organization. They’re focused on building technology internally and strategically leveraging the tech already being provided by vendors. Other law firms and organizations may not have dedicated AI teams, but they have certain innovative lawyers or members of their teams who are tech savvy, entrepreneurial, tuned into the latest market developments, and importantly, acutely aware of the needs of their organization. And those impressive self-starters are themselves initiating efforts to build new technologies that are both highly innovative and closely tailored to the needs of their organization.
Cheat sheet: Here’s what the tech can already do (and the list is growing)
To get a sense of the use cases already available, I’ve noted some leading tasks that gen AI is pretty good at already—and increasingly improving. This list below is growing quickly, and while your specific use case may not be listed, that doesn’t mean it doesn’t already exist (more likely, it does exist—the list is just a sampling). A suggested path forward is to survey the list, survey what’s available (via internal development or outside offerings), and quickly begin to pilot, test, and iterate from there.
Growing list of generative AI tasks for legal services:
- drafting specific clauses and generating insights (generating suggested clauses to include in a contract)
- drafting first drafts of legal documents (creating a first draft of an agreement)
- marking up drafts (creating initial markups or “redlines” of drafts received)
- summarizing documents (uploading a number of documents and having the tech summarize them quickly)
- reviewing (for example, using chatbots that support review of documents throughout the drafting and reviewing life cycles)
- translating (translating texts into virtually any language with a tailored degree of accuracy to legal jargon and meaning than more standard translation services)
- searching and retrieving (pointing to a database and finding work product and information easily)
- comparing (comparing documents at scale is possible and easier than manual comparison)
- responding to case-specific questions (with case law citations)
- deposition outline automation
- e-discovery automation
- document reviewing at scale
- IP management automation (including generating patent applications and identifying potential infringement)
- performing due diligence at scale
Next up: Evaluate your data readiness
“Data is the new oil,” said the British mathematician Clive Humby back in 2006; and this bite-size wisdom has been rightfully expanded upon to show the importance of refinement for both; in order to be valuable, data, like oil, ought to be refined, cleaned, and structured into an ingestible and useful format.
For any legal organization, your structured and unstructured data sources will be the oil that powers the engines of any gen AI technology ultimately used. Structured data sources might include your document management system (that is, any system you use to store your signed agreements and drafts) and any playbooks or sources of drafting notes and guidance. Unstructured data sources can include written notes, writings, and documents stored but not organized with a specific theme or purpose—essentially anything that can be analyzed by the technology and may be useful for its training, but that’s not grouped with any broader set of tooling, documents, or taxonomy. The more data you have as an organization, the better positioned you are for success when considering gen AI legal tech adoption, because your training journey for the tooling will be that much smoother. Simply put, better inputs (organizational data) lead to better outputs (results of the gen AI legal tech).
And then: Are you in the cloud yet?
If you’re not in the cloud yet, there’s no better time to fly. Importantly, for adoption of gen AI, your back-end technology stack (meaning the technology infrastructure powering your law firm or organization) should be cloud-supported (versus being “on premise”), because without being cloud-supported, the gen AI cannot operate as smoothly, efficiently, or securely.
Pro tip: Before thinking about adopting legal tech, check in with your favorite internal IT friend, and ask a few questions about whether you’re already “in the cloud” as a firm or department, and if not, what your data storage looks like. If you’re not already cloud-supported, consider kicking off a parallel process to become cloud-enabled as you’re evaluating legal tech to build or adopt.
The leading question: Consider the right legal tech vendor or internal-build approach for you
So you’ve figured out the leading use cases you think will bring real impact to your organization. Keep in mind through this exercise that most legal tech offerings are open to and amenable to iteration (particularly at this relatively early period). Vendors will likely have off-the-shelf products available, but they’re generally open to and happy to iterate with legal teams and tailor their offerings to the specific use cases you’ve identified.
Keep in mind that iteration means you’ll likely have an initial pilot period where you work with the vendor directly to test out tech and tailor it to the use cases, so you’ll need a dedicated internal team that is ready to test and practice real-world application of what’s provided and a dedicated team lead who can “project manage” the workflow and pilot period, interface with the legal tech vendor team, and effectively lead the development and tailoring of that legal tech. This is where strategic direction and guidance (understanding the goals and needs of your organization), ability to lead a team toward those goals, and a set definition of what success looks like is a real difference maker. For law firms and legal departments, leadership of the legal tech pilot workstream might vary across a tech, knowledge management, legal ops, or AI team and will generally be overseen by, or done in collaboration with, the organization’s innovation lead.
Legal tech pilots will range in their time frame, but a good place to start is one to three months. Licensing considerations of the fulsome tech from there vary too, but at this stage of rapid development, shorter commitments may benefit the users (for example, a six- to 12-month initial commitment), given the tech and needs are quickly developing.
Finally, the adoption: Organizational psychology for the win
“Simplicity is the ultimate sophistication.” – Steve Jobs
Any tech that’s ultimately chosen will be new for many in both concept and practice. It ought to be easy and intuitive, and something to look out for when reviewing the tech is the ease and comfort of using it. Is there a single search bar or entry for text and a minimalist design with clear instruction?
The goal is to find something that can be used as an easy copiloting tool, something that will be open on your laptop all day (like email), where you’re constantly referencing it and using it for your communications and workflow. Ease and simplicity may be the greatest currency when it comes to the adoption (and, ultimately, the success) of any given legal tech, especially something so new.
Tailored training, too, is critically important. Alongside ideation and strategy mapping around legal tech build or vendor selection, it’s a good idea to start thinking about the training and upskilling needed across users (that is, lawyers) and what they’ll need to learn to maximize their use of the tech. Training on prompt engineering and design thinking (described in my earlier article on legal tech and next-gen lawyering), and detailed explanations of how the tech can be used, would greatly benefit lawyers, particularly those without exposure to gen AI technologies and who have largely been classically legally trained.
It’s all doable. We’re right here doing the same thing too!
At McKinsey Legal, we’re working through all the same considerations, and the suggestions here are based on our own innovation transformation journey over the course of this year. We set out a number of pillars for innovation at the start of 2023, with an architecture anchored in thought leadership, client centricity, legal tech, and our people mission. Each of these pillars had a number of initiatives mapped to the identified goals of our department. And we quickly realized that wrapping our arms around the latest developments in generative AI, and how they applied to our goals, needs, and most importantly, the “2.0” of our department, was critical to jump on and get right. The suggestions above are based on our learnings so far, with the recognition that we’re only getting started.
In the words of Coach Taylor from one of my favorite series of all time, Friday Night Lights: “Clear eyes; full hearts; can’t lose.” In my humble opinion, we at McKinsey Legal have an extraordinary department full of distinctively talented lawyers. To be able to set us up for greater enablement, expertise, and acceleration by automating work flows and leaving room for our greatest, most creative, highest-value work feels like a poignant privilege. Onward and upward we go as we explore, pilot, and adopt legal tech.