As in-house counsel continue to face cost pressures from within their organizations, external counsel need to help respond to those pressures. In our discussions with clients about solutions, the conversation inevitably focuses on leveraging technology to reduce costs while improving efficiency. Clients regularly ask about AI and how to use machine learning to save money. One clear answer is when responding to eDiscovery requests.
EDiscovery is a pain point for in-house counsel, who often hold the perception that it is both disruptive to business and expensive. However, incorporating AI/machine learning tools into the eDiscovery process allows eDiscovery practitioners to deploy technology that significantly reduces the cost and the pain.
Historically, 70-80% of the cost of an eDiscovery exercise related to the lawyer review phase. While there were always costs associated with collecting, processing and loading data into a review platform, the bulk of the expense generally incurred in this phase. As data volumes grew, and we collected terabytes instead of gigabytes, the problem exacerbated. In cases where clients delivered hard drives containing hundreds of gigabytes or terabytes of data, the potential cost for the review became significant.
Machine learning has largely resolved this big data problem.
How It Works
Once the data is collected and an early case assessment and analysis is complete, the data is loaded into the review platform and each record is categorized by the system. As lawyers begin the review, their decisions create a model that incorporates algorithms to train the computer as to which records are relevant.
Machine learning uses techniques such as text analytics and pattern recognition to develop its rules. It creates a relevance model that it uses to automatically categorize records, serving up the most relevant records for review first. During the process, the system reevaluates and updates frequently, allowing supplemental records to be added to the review set and ensuring its own ongoing training.
Once the machine no longer offers up likely relevant records, the remaining records are sampled and validated to give the clients and courts certainty that the process is defensible. In lieu of a linear review after keywords were applied, we no longer need to know the keywords, the secret code names for projects or other nuances of a case in advance.
The lawyer time is spent focusing on key documents instead of the mundane emails, which are now automatically excluded from the review. The focus is on the litigation strategy, not the volume of data.
For in-house counsel with large volumes of litigation or huge volumes of data, it is imperative that this technology be adopted. Let’s use as an example a case where 1,000,000 records are loaded into the review platform after the data has been deduplicated, date filtered and email threaded. The client believes only 20,000-30,000 records will be relevant, so the challenge is to identify those 20,000 records without reviewing all 1,000,000 records. Machine learning is the solution.
In recent experience, lawyers are only reviewing 50,000 records out of the 1,000,000 records to find the 20,000 that are relevant; the remaining 950,000 records never need eyes-on review. Imagine the cost reduction and time savings of a review of 50,000 records compared to the time prior to machine learning, when the review budget was for 1,000,000 records.
EDiscovery teams are successfully using machine learning to only review small subsets of massive volumes of data. Quickly identifying the records that are required for litigation, investigations and regulatory production obligations is key.
While these systems come with a significant investment in technology, the people and the process to deploy the technology is equally critical. A team of eDiscovery practitioners who bring people, process and technology to the table always provides the best solution.
Susan Wortzman is a partner in the Toronto offices of McCarthy Tétrault and leads the firm’s e-Discovery and information management practice. She is also the founder of MT>3, formerly Wortzmans, Canada’s leading law firm specializing in providing advice and services relating to the management of digital information, which was acquired by McCarthy Tétrault in 2017.