When Judge Andrew Peck published the legal opinion that rocked the eDiscovery-verse, Da Silva Moore, over a decade ago I was convinced that the future for the practice of law would be inextricably intertwined with AI and machine learning. And yet, here we sit today, still having to dispel the false narrative of human review as the gold standard in eDiscovery. What went so wrong? And how do we begin tipping the scales in favor of AI adoption in eDiscovery?
A few critical missteps in the launch of some AI legal tech startups are to blame for the lack of adoption by lawyers and organizations alike. Savvy organizations are flipping the script on these missteps and seeing AI software in eDiscovery embraced at faster rates as a result.
The most well-known early iteration of machine learning or AI in eDiscovery, Predictive Coding, hit the stage in a big way. Even going so far as patenting the term and leaning into being the first AI solution for legal. In an industry predicated on risk mitigation and not often welcoming to change (heck much of legal precedent dates back to before there were typewriters let alone AI), being the newest cutting-edge technology is not necessarily a selling point.
In reality, the technology powering TAR, active learning, and yes, predictive coding, was far from being novel outside of legal. Machine learning and artificial intelligence were first conceived in the 1950s with many flavors widely in use by industries including financial services, retail, manufacturing, healthcare, and more. While AI, machine learning, and natural language processing (NLP) may have been novel in the legal industry, they were tested and trusted by a wide array of industries at large.
A second critical misstep with early applications of machine learning and artificial intelligence in legal technology companies was a lack of transparency in terms of how the technology actually worked. By hiding the secret sauce, and in many cases pretending it was much more advanced and sophisticated than it really was, early eDiscovery AI became off-putting to law firms and legal practitioners. No lawyer worth their salt wanted to have to explain to a judge how this black box worked. It was hard to establish trust in a solution or process that was unnecessarily complicated.
Newer iterations of AI-powered legal technology and especially eDiscovery tools have embraced a more transparent approach. Rather than shrouding machine learning in lambda calculus and advanced statistics (hello F score, precision, and recall), the next-gen of automation solutions use simpler language and seamless integration to appeal to a wider user base. This approach, combined with the proliferation of AI-powered technology in people’s personal lives (Google, Uber, GPS, and even spam filters) has served to greatly demystify legal AI.
Another thing that discouraged law firms and legal service practitioners from embracing legal AI was the need for both a unique “AI Workflow” and an over reliance on statisticians, AI experts, and linguists to execute. In making eDiscovery AI appear sophisticated and complex, early tools actually discouraged many legal practices from adopting the technology. They were worried they lacked the skills to effectively leverage the tech and frankly were afraid of making an error that could dramatically impact their case.
Now, much in the way that Google utilizes arguably the most sophisticated search algorithms while it is so simple to use that my 7-year-old nephew can type in a query, newer Legal AI solutions have begun leaning into iPhone-easy user interfaces and seamless AI integration. Coupled with the thousands of cases that have benefited from greater accuracy and speed, not to mention reduced cost, of uncovering evidence or insights using AI, the shift away from over-complicated solutions requiring special skills, workflows and expertise have tipped the scale in favor of greater AI adoption.
The lingering misperception that human “eyes on every legal document” approach to document review is somehow a more accurate “gold standard” deterred many veteran legal professionals from dipping their toes into using AI or machine learning-powered solutions in their legal work.
Studies have consistently shown that human review lags substantially behind the 98%+ accuracy boasted by most major AI-powered eDiscovery tools. From the NIST TREC legal Track that published studies from 2006-2021 to the more recent bake-off by Lawgeex, to actual user experiences over the last decade, the resounding conclusion is that humans are the gold standard mentality is fundamentally flawed.
A final mitigating factor to broader artificial intelligence adoption in the legal sector boils down to plain old fear. Lawyers and legal professionals generally are a fairly risk-averse group. The combination of overly complex information technology and a misplaced belief that human review was somehow better created a perfect storm of perceived risk. Because AI-powered solutions worked faster and often saved substantial amounts of money on a matter, many legal teams incorrectly assumed that there must be a greater risk in relying on algorithms to categorize, analyze or remove data from human review.
As noted above, the over-complicated and expert-driven workflows of legacy legal AI solutions have been replaced with more intuitive and easy to explain ones. Countless studies and cases have shown not just parity with human review, but results that far surpass human accuracy and recall. As a result, the risk calculus has begun to shift somewhat in the last few years, to one that places manual linear review squarely in the riskier category. Some legal sector thought leaders have even contemplated a future where opting not to use available technology to uncover insights, evidence or context might subject a legal professional to scrutiny under Model rule 1.1 or other state bar requirements of technical competency.
In addition to the evolution in the perceived value of AI-powered legal technology solutions and the fairly dramatic shift by bar associations to include technical competency as an ethical obligation for lawyers, a big part of the shift to AI acceptance has come from the legal tech companies themselves. Here are a few ways legal tech software companies (like none other than Reveal-Brainspace) are tipping the scale in favor of broad legal AI adoption.
Brainspace led the charge in developing eDiscovery tools that took AI-powered insights to the masses by creating visualizations that made sense to technology experts and non-experts alike. With intuitive and visually engaging ways for legal practitioners to streamline insights driven by AI and machine learning, Brainspace made using AI less daunting of an undertaking.
The original AI in eDiscovery, predictive coding, required lots of planning and an alternative workflow any time a legal professional wanted to deploy AI. Newer iterations have made applying AI to a data set as simple as toggling on a switch, and are built to work seamlessly in tandem with traditional review. You can now access insights (like communication patterns or key concepts) without having to code at all. As a result, the barrier for a law firm to employ AI is far lower today.
The first applications of AI in law were one-trick ponies, with rigid and often singular applications. Today there is an entire spectrum of AI and machine learning that can be applied to a legal issue jointly or separately to dramatically reduce time to insight.
Additionally, advancements in the sophistication of AI-powered technology now allow for legal departments to browse an AI model library in much the same way they would scroll through their Netflix queue. But, instead of finding the next rom-com to binge, they can find the perfect model to apply to their big data set to dramatically accelerate time to evidence.
Today’s legal teams have a higher expectation in terms of ease of use and intuitiveness of their legal tech tools because in their daily life applications and tech have leveled up substantially. To meet increased expectations from clients, platforms have seamlessly integrated AI into the fabric of their legal research. It is now easier to access AI-driven insights from relevant case law and apply them across a matter, or even a portfolio of cases. Long gone are the need to perform complex statistical analysis, in favor of intuitive and easy to interpret results in just a few clicks.
Between the improvements in next-gen legal AI, the familiarity of AI in our personal lives, and the mountain of proof that AI finds the information you need faster, the future of the legal market looks bright for increasing AI adoption.
I also am of the opinion that a healthy dose of “Fear of missing out” (FOMO) will serve to massively accelerate adoption in the coming years. AI-powered eDiscovery and legal technology helps lawyers practice law better. The results speak for themselves. Savvy legal professionals looking to get ahead and keep discerning clients happy will continue to gravitate to AI-powered legal tech AI companies to do so.