Corporate legal departments and law firms are becoming well-versed in the benefits of investing in technology. Many in the industry know the cost savings and other efficiencies that Gen AI and other advanced technologies bring. For many legal teams, utilizing technology isn’t a ‘nice to have’. The growing digital evidence available to legal teams makes Technology Assisted Review (TAR) and other advanced technologies necessary to wade through the volume of data involved in the eDiscovery process. In recent years, we have seen legal departments and law firms investing in sophisticated and complex technologies, only for many of them to subsequently fail to leverage what these tools have to offer. This is often because they have not implemented the technology correctly or onboarding and training have been insufficient.
It is also uncommon to dig further into a technology’s capabilities, which are already creating efficiencies, to get an extra 10 percent return. I often equate it to having the power of a Ferrari behind you but deciding to use it for the school run. However, considering the cost of large legal projects and how much 10 percent could equate to, it is certainly worth the extra investment of time and effort. Here, we will explore how legal professionals can go beyond what the technology offers out of the box and customize their approach to supercharge results, using TAR as an example, as it is becoming more ubiquitous in the market.
Turning to advanced analytics to drive efficiencies and savings is the key to achieving improved results. With TAR technologies, this means inserting customized features into predictive coding workflows to achieve superior recall and precision much sooner in the process, beyond what would be achievable with the tool’s out-of-the-box capabilities. An experienced team will cull documents and reduce the number sent to human review while providing transparency into results through clear dashboards and insights. Legal teams who don’t programmatically measure and track the effectiveness of using TAR are leaving money on the table.
With so many TAR tools on the market, we recently evaluated the performance of three popular, commercially available offerings, including Relativity’s Active Learning (RAL), Reveal’s Brainspace (BRS), and Reveal AI (RAI, formerly NexLP), to identify a clear winner if such a distinction existed.
Our analysis included results when using each tool’s flagship queue to select the appropriate training documents, compared with the results when a human consultant selected training documents based on the algorithms’ supervised learning selections. Although one of the three tools we tested performed significantly better than the other two, we discovered something far more interesting: human intervention during the training, specifically the selection of training documents, improved model results across the board, surpassing even the best result without human intervention.
This is because humans can synthesize the importance of training to a target result, whereas the tools select documents based on pre-built goals that may only be effective in some reviews. It’s important to consider that machine learning should not be relegated entirely to the machine. After all, active learning utilizes supervised learning, with SMEs reviewing documents to train the machine to classify responsive and non-responsive documents . We should not ignore the performance gain from human analysis of the active learning output and resulting human input to identify more effective training documents for machine learning. Machine learning represents an opportunity to augment and accelerate human analysis, not replace it.
The key to boosting results with technology is often not relying on the technology straight out of the box, but introducing a qualified human to drive an extra 10 percent of efficiency, which can mean saving tens if not hundreds of thousands of dollars. In fact, the most important criterion is not the selection and use of the tool but the human overseeing the workflow.
Russell Hutchins is Vice President of Advanced Technologies at Epiq.
