Artificial intelligence, or AI, is a tech buzzword that’s been in the news a lot. Legal tech is no exception, and it feels like every company is saying they sell software that leverages AI. The promise is that AI will make work effortless by automating tasks and eliminating the need for manual, human work. It’s a sexy proposition, but what’s the reality?
AI certainly holds great promise for helping the legal profession and has a lot of value to add today, but the reality is that there is a long way to go before technology develops to a point where this promise can be fully realized. Unfortunately, the way AI is often marketed in legal tech implies we’ve already arrived at that destination, and that has impacted the expectations of some legal teams and how they think about purchasing legal technology.
To illustrate this, let’s compare how AI is portrayed in another industry where AI is becoming increasingly prominent: the auto industry.
Shifting from Manual to Automatic.
The prospect of self-driving cars arriving “in the next few years” is exciting. Anyone who commutes in a car will appreciate being able to replace the monotonous, daily routine of driving in stop-start traffic with the ability to do, well, almost anything else.
Today, no car manufacturer claims that their AI technology is ready to do it all. This doesn’t stop them from marketing the automation already in their cars, like adaptive cruise control, auto-park, and voice control. They’re not full-blown AI, but they do make driving easier and more enjoyable. Your workload is reduced, and your role changes from a worker performing the tasks, to a supervisor monitoring that the car is doing the right things.
In contrast, many legal tech solutions now lead with AI at the center of their marketing. Understandably, many legal teams get the impression that AI must be central to legal tech, and it will replace the need for any review or supervision by a lawyer. Unfortunately, that doesn’t reflect reality. Getting AI up to a standard where it could reliably replace a legal professional is a really, really tough technical problem, and one that won’t be solved any time soon. (Even worse, a recent survey found that 40% of AI startup firms really aren’t that intelligent, containing no “real” AI tech.)
To understand why it’s such a tough problem, let’s look at one of the major challenges facing AI: ambiguity. For self-driving cars, Tesla CEO Elon Musk has said: “[There is] no problem recognizing stop signs and traffic lights. But you do get ambiguity in some complex intersections with traffic lights. Like, which one’s the right light to focus on? Even if you’re a person, it’s not always clear.”
The same type of problem exists for legal AI. Take, for instance, AI that is tasked with interpreting a contract and identifying its governing law. On the surface, the solution may appear relatively straightforward — just look for language that says something similar to: “this agreement is governed by the laws of California.”
But what about agreements with conditional governing law clauses? “If the Customer is purchasing from a Vendor entity located in the United States, Canada, or Latin America, this agreement is governed by the laws of California. Otherwise, this agreement is governed by the laws of the jurisdiction in which the Vendor entity is organized.” Suddenly the problem becomes much more difficult.
Lawyers can still disagree on the interpretation of a clause, even after years of legal education and experience. If highly skilled humans can’t agree, how can AI decipher all the different permutations of language that are even more numerous and confusing than the “complex intersections” we encounter when driving?
The answer is that it can’t — at least, not today. Technology is simply not advanced enough to achieve anything close to 100% accuracy.
Unlike self-driving technology, the worst consequence of flawed legal AI is unlikely to be death, but the consequences can still be serious. For a profession that places a premium on accuracy, even a 95% accurate rate may mean a human still needs to review all of the work AI has done.
How You Should View Legal AI
In contrast to the auto industry, we think legal tech has things a bit backwards: overplaying the capabilities of legal AI, while underplaying other complementary technologies that are just as good, if not better, at improving efficiency. The danger with this is that it leads businesses to have high expectations about the capability of AI — expectations which inevitably lead to disappointment: “Why does the software make these mistakes?”
Yet, there is value to AI even if it’s far from perfect today. While it may sound disappointing if AI produces a task accuracy rate of 80%, that’s still saving you about 48 minutes of work per hour! From a practical perspective, the good news is that even though AI won’t let you take your hands completely off the steering wheel, it is very good at playing a supporting role in legal tech by performing simple tasks with a high degree of accuracy.
By viewing AI through the right lens —as an assistive, but imperfect, technology — it helps to set the right expectations.
When evaluating legal tech solutions, remember your ultimate goal: to become more efficient, because there simply aren’t enough hours in the day, and to better serve your clients. So instead of focusing on AI accuracy rates, ask questions like:
- How much time will I save by using this product? How much of that relies on AI to deliver those time savings? What other features does the product offer that save time?
- Will I be able to accomplish more in my job by using this product, than by not using it?
- Will this product change the way I like to work, and will that create more work for me or less?
(Also consider that while a 90% accuracy rate sounds substantially better than 80%, it’s actually only a marginal improvement of 6 minutes per hour over the 48 minutes per hour that 80% offers.)
Making Progress Today
Most people haven’t delayed buying cars because self-driving technology isn’t ready. They simply decide how much assistive technology they want and accept that human supervision and judgement will still be better than the manual alternative.
The same can be said for legal tech — if your goal is to become more efficient and productive, then anything that helps to achieve that is a win. So consider AI, but focus on more important things, like how intuitive a system’s user interface is, or whether it’s actually making you do more work than the manual process it replaces.
Even though nearly-perfect AI doesn’t exist today, that doesn’t mean that you need to wait before implementing a solution that can free up more hours in the day. Imagine that — a work environment where you can use technology to automate the mundane, allowing you to focus on the strategic. That’s a much more sexy proposition to us.