Know your toolset: a guide to AI-powered eDiscovery

Recent advances in legal technology are fundamentally reshaping the way lawyers conduct investigations, discovery and case preparation. Tasks that once required large teams of manual reviewers are now being carried out by sophisticated and rapidly developing AI-driven tools.

Modern platforms like RelativityOne bring together a suite of distinct, purpose-built AI tools, each addressing a different aspect of the legal process. Some give an overview of, or identify relevant documents within, large and complex data sets. Some extract and analyze key information including facts or metadata. And others help to identify and protect sensitive information before production.

The rapid evolution of this field is often daunting for a legal team at the outset of an eDiscovery process. Key to success: understanding the various tools’ capabilities, limitations, best-practice applications, and how they integrate with existing eDiscovery workflows. Experienced eDiscovery professionals help lawyers navigate the complexity, aligning the most appropriate solutions and workflows to the specific demands of each matter. 

Understanding your data and gaining case insights early

Tools that accelerate the legal team’s understanding of its dataset are particularly valuable during Early Case Assessment (ECA) in a matter’s initial stages. More established approaches like conceptual search and clustering use indexes to identify or group documents by theme or topic. 

Recently, generative AI is extending these capabilities through natural language interaction, aimed at operating like a highly efficient research assistant. Relativity’s aiR Assist allow lawyers to ask conversational questions (like “are there any documents suggesting potential bribery?”) and receive concise responses drawing upon and distilled from a large underlying dataset, with cross-links to specific documents—producing accessible insights on-demand.

These capabilities remain valuable throughout a matter. As case theories are developed, they can be used to test assumptions and preliminary findings. And they are powerful during depositions and witness preparation where rapid access to relevant evidence is critical.

Imminently, Relativity will release an AI-powered ECA dashboard which automatically extracts from a dataset key topics and themes, people and relationships, and trends. This accelerates lawyers’ identification of emerging themes and potential connections, overall understanding of a dataset, and informed decision-making.

Locating relevant documents within your dataset

Human document review is resource-intensive and costly. As the volume of data grows exponentially, it is increasingly important to defensibly identify manageable data subsets for human review. 

Keyword searching remains useful here, but AI-based tools like Relativity’s aiR for Review offer significant improvements, frequently delivering consistent and precise relevance and issue predictions at scale and often over 200 times faster than human review. (They can also operate 24 hours a day.)

Generative AI can give richer insights beyond binary relevance and issue predictions. For each document, aiR for Review can give summaries, explain its coding predictions (including giving counterpoints to them) and identify supporting excerpts. Often such insights are not commissioned during manual reviews due to the material impact on time and cost.

These latest technologies complement traditional machine learning such as continuous active learning (CAL), which continuously reprioritizes for earlier human review those documents most likely to be relevant, based on previous coding decisions. For example, a population of documents that generative AI predicts to be relevant can be further prioritized with CAL, surfacing the most critical material for second-level human review earlier, and where statistically defensible, further reducing the overall review population.

Extracting information and generating content from documents

Once relevant documents are identified, it is important to understand what those documents mean in context.

Determining the precise sequence of events is often critical. Historically, chronologies have involved laborious manual work. Generative AI like Relativity’s aiR for Case Strategy automates the process—extracting facts to generate timelines which can be presented visually, filtered dynamically, and refined as the matter evolves. 

It goes one step further: Generative AI can run over the extracted facts to generate first drafts of witness and interview outlines efficiently. These capabilities will soon extend to allowing lawyers to generate tailored work product like case summaries or issues analysis using natural language prompts.

Similarly, other AI features (also included in Relativity aiR for Review) facilitate custom analysis across a document population, extracting specific information at scale even where it appears in unstructured or varied formats. The results are aggregated in a structured output that can be filtered, sorted, interrogated and exported, enabling legal teams to move quickly to meaningful analysis. Historically, this was difficult to achieve without extensive manual effort or rigid pattern-based approaches. For example, legal teams may use custom analysis across thousands of contracts to identify and analyze references to delivery dates, pricing terms, or performance milestones.

Identifying and protecting sensitive information

eDiscovery frequently involves producing documents containing sensitive personal information or proprietary business data. Inadvertent disclosures give rise to commercial and reputational risks. Further, various legal obligations may apply, including during litigation and investigations—for example the EU General Data Protection Regulation (GDPR) for certain personally identifiable information (PII) and protective orders in the US for confidential business information (CBI). 

Historically, the identification of sensitive information relied heavily on keyword searches and pattern-based approaches. For PII, predefined search term lists or regular expressions often had high false positive rates alongside the risk of false negatives. CBI identification frequently relied on human review, given that nuance and context are critical. Redaction itself could be automated somewhat, yet determining what material required redaction was time-intensive and often subjective.

Modern capabilities are a significant evolution here. They represent a step-change improvement in proactivity, scalability, defensibility and the mitigating of inadvertent disclosure risk particularly in large-scale reviews. Generative AI, pattern recognition and keyword-based approaches when used together can surface a wide spectrum of sensitive information, from structured identifiers (like driver’s license numbers) to more nuanced material like trade secrets—then escalate it for review and handling, at the same time providing transparency and facilitating quality control by citing supporting excerpts and giving reasoning.

Platforms like Relativity’s aiR for Review are offering dedicated analysis types for both PII and CBI. In due course these technologies are expected to further streamline workflows by integrating automated redactions directly into the analysis process.

Using new technologies insightfully

The evolution of legal technology is transforming litigation and investigations, introducing powerful tools that can accelerate and enhance every stage of the process. Yet this comes with technical complexity. Success depends not only on access to the technology, but also on applying it strategically and defensibly, aligned with the specific needs, nuances, and risks of each matter.

Here, expert eDiscovery teams play a critical role. They guide tool selection, design tailored workflows and ensure rigorous validation. Closely collaborating with legal teams, they help to transform document review from a reactive, labor-intensive exercise into a more strategic, insight-driven process.

This post reflects an ongoing journey of applying AI in eDiscovery. Stay tuned for our next blog post which will consider when AI should and shouldn’t be used, and how AI works alongside established eDiscovery technologies to enhance legal workflows.

About A&O Shearman’s eDiscovery capability

The A&O Shearman eDiscovery team is one of the most highly certified and experienced in the industry, with ten Relativity Masters—the highest level of accreditation available. Our expertise is grounded in more than a decade of hands on use of Relativity, during which we have worked closely with the Relativity team to refine workflows, provide structured product feedback, and support the platform’s ongoing development.

This longstanding relationship gives us early access to the latest AI capabilities, enabling us to introduce new features to client investigations as soon as they are released. By pairing these innovations with rigorous quality controls and the judgment of seasoned specialists, we help clients surface key facts quickly, manage exponential data growth effectively, and navigate investigative challenges with confidence and strategic focus.

To learn more about the services we offer, please contact Christina Zachariasen, Daniel Smith or Natalie Lau.