AmLaw 50 firm and Control Risks leverage aiR for Review to shorten first-pass review from weeks to days, achieving cost savings for a multinational client.
A faster, cheaper and better data review:
- 80% reduction in overall review time
- First-level review completed in four days
- Over 50% in first-level review cost savings
A sensitive internal investigation
An AmLaw 50 firm was retained by the Audit Committee of the Board of Directors of a multinational, US-listed corporation with significant operations in Latin America to conduct a sensitive internal investigation involving alleged violations of the Foreign Corrupt Practices Act (“FCPA”), fraud and other possible misconduct.
Counsel retained Control Risks, a RelativityOne Gold Partner, to assist with the investigation, including support for digital forensics, eDiscovery, structured data analytics, and forensic accounting.
The investigative team was tasked with reviewing more than 120,000 predominantly Spanish-language documents, conducting over a dozen interviews, and presenting its findings to the Audit Committee and the client’s external auditors within a strict four-week deadline.
Accelerating fact-finding with generative AI
To meet this timeline, Counsel and Control Risks recommended leveraging Relativity’s aiR for Review instead of traditional human first-pass document review. Using aiR for Review reduced first-level review time by ~80% and helped achieve ~50% in cost savings, significantly improving efficiency and bolstering client confidence in counsel’s ability to meet the deadline.
Counsel and Control Risks collaborated to explain the operation, reliability and benefits of aiR for Review to the client as well as its external auditors. With the client’s approval and assent of its external auditors, the team worked together to develop a tailored workflow to address the unique demands of the project:
Custom prompt development
Counsel and Control Risks collaborated to develop two aiR for Review prompts. The two aiR for Review prompts used were (a) “Relevance” and “Key Documents”, which evaluates whether a document is pertinent to the case and if it should be considered “key” and (b) “Issues”, which involves a more specialized analysis focusing on whether documents fall into a specific category or topic.
Similar to a conventional document review protocol, the prompts included a detailed matter overview, an explanation of key individuals and entities, and a summary of the underlying allegations and relevant laws, regulations and compliance policies.
The prompts were refined following a sample batch review, which included three sample sets with 400 files per batch that were tagged by aiR for Relativity and then subject to human review. This initial calibration process – which was completed in two days – ensured that the AI tool was capable of accurately and consistently applying the correct tags.
AI deployment
Once the prompts were finalized, Control Risks deployed the aiR for Review tool to analyze over 120,000 documents and identify relevant, key and issue-specific files.
- The 120,000 documents used in the aiR for Review analysis were the final results after applying email threading and search terms.
- The documents were split into multiple tranches, each run through aiR for Review as a separate exercise to return results quickly and then batched out to the review team on a rolling basis.
- Throughout the review, the team ran sample analyses across certain of the relevant documents and batched those out to the review teams to test that the model was working as expected.
Second-level review
A second-level review workflow was established to prioritize documents based on their aiR for Review predictions, pushing those predicted to be most relevant to the top of the review queue.
Running aiR for Review to identify specific issues allowed us to understand the volume of files for each issue and tailor second-level review batches based on our prioritization of the issues.
aiR for Review’s output, including rationales, citations and links to the relevant text, provides greater detail and consistency than traditional first-level reviews. This sped up the second-level review as well as the interview exhibit selection process.
aiR for Review also leverages generative AI to give written rationales for each document, explaining its tagging decision and allowing second-level reviewers to quickly understand the basis for the initial assessment.
Foreign language optimization
aiR for Review has multilingual capabilities, including for Spanish language reviews and those containing a combination of languages.
Our review was predominantly in Spanish, and aiR for Review was instrumental in performing the first-pass review quickly – and at a reduced cost.
This approach allowed us to proceed with a smaller bilingual team of first-level reviewers primarily responsible for a quality control assessment of documents predicted to be “Not Relevant” by aiR for Review.
A high-quality outcome tailored to the client’s needs
The case team leveraged aiR for Review’s insights to:
- Complete their second-level review with targeted precision over 85%.
- Identify interview exhibits efficiently.
- Meet a tight reporting deadline with confidence.
Control Risks was engaged to provide additional support beyond the digital forensics and eDiscovery needs of the investigation. With aiR for Review's ability to identify relevant documents in a time-efficient manner with increased accuracy, we could perform parallel structured data analytics and transaction testing workstreams, allowing us to do a deep-dive to examine the financial datasets much more efficiently and holistically.
By combining advanced AI capabilities with human expertise, Counsel and Control Risks delivered a high-quality outcome tailored to the client’s needs. Together, we completed the review expeditiously and to a standard that satisfied the Audit Committee and the client’s external auditor.
This matter showcased how aiR for Review’s ability to decipher foreign language documents and provide actionable insights can transform demanding document review processes and fact-finding.
“We are thrilled to be leveraging Relativity’s aiR for Review, which is already realizing its massive potential to transform how we uncover facts in disputes and investigations—faster, cheaper, and better (greater accuracy and defensibility). Traditionally, you could only achieve two of those three goals at any given time, while significantly sacrificing the third, but with aiR for Review, you can truly have all three. It’s an exciting paradigm shift for our industry, and we’re eager to see the impact generative AI will have on our ability to deliver exceptional results for our clients.”
— Brad Kolacinski, Partner and Global Head of eDiscovery, Control Risks
“aiR for Review enhances efficiency in the review process, particularly in foreign language reviews. It simplifies the understanding and analysis of foreign language documents, helping to lower costs associated with translation workflows and foreign language reviews.”
— Maggie Burtoft, Principal and Head of eDiscovery Project Management in the Americas, Control Risks
“The implementation and use of aiR for Review was intuitive, allowing our eDiscovery experts to easily create aiR for Review projects to run concurrently through the application. The yielding results were accurate and concise, allowing the case team to review with speed in order to meet a tight deadline.”
— Zehra A. Yazgan, Director and Senior Project Manager in the US, Control Risks