AI Case Study
Muckle LLP reduced manual document review workload by 95% using used technology assisted review
Muckle LLP engaged Advanced Discovery's eReview services to automate the analysis of a large number of documents for an upcoming breach of warranty case, making discovery of pertinent information quicker and more efficient. It was estimated that 95% od the documents for review were automatically tagged as not relevant, reducing the manual review workload.
"Technology Assisted Review (TAR) is a process that takes the review input generated by a subject matter expert (SME) and uses this to formulate coding decisions across a universe of documents. Often referred to as “predictive coding,” this process informs an automated workflow to categorise documents within the data set as either “relevant” or “not relevant” without human intervention. Additional quality control processes are integrated into the review to ensure that false negative and positives are avoided. Through several rounds of training, TAR began to “learn” from the coding decisions made by the Muckle experts. The “training rounds” were then validated through quality check or QC rounds. Working very closely with Advanced Discovery’s TAR experts, inconsistencies were identified and rectified in an iterative process designed to help achieve stability within the system."
"Muckle LLP, a commercial law firm based in Newcastle upon Tyne, in the North East of England, engaged Advanced Discovery as their eDiscovery partner in a breach of warranty dispute for an aerospace client. One million documents and a looming deadline for production meant that the team needed expert advice on how to move quickly, without compromising quality or taxing their client’s resources... The high density of complex contractual documents and financial models and spreadsheets was daunting – especially given our tight deadline.”
"Upon completion of the TAR project, approximately 35,000 documents from the relevant data set of 660,000 – were identified as likely to be responsive. That meant 95% of the documents were defensibly excluded from the manual review process. The Muckle team was prepared for disclosure two weeks ahead of their deadline."
Began with 660,000 documents including spreadsheets, contracts etc. "Before identifying TAR compatible documents for the workflow training process, Advanced Discovery’s Relativity experts began culling the data using near duplicate analysis, domain parsing and concept clustering to remove known non-responsive file types, such as advertising and spam, typically cluttering email inboxes. These documents were subjected to additional analyses and quality control steps to confirm their lack of relevancy. 340,000 documents were excluded from the document set.
From the data identified for inclusion in the TAR workflow, a representative sample of documents were batched out for the initial training of the software.