As data volumes continue to grow and litigation timelines tighten, legal teams are under increasing pressure to review documents faster, more accurately, and cost effectively. Over the last decade or longer, Technology-Assisted Review (TAR) tools have become one of the most effective methods to meet these demands—but it is also one of the most misunderstood.
For many legal teams, understanding what TAR is, how it works, and when it makes sense to use it is essential to building confidence in modern eDiscovery strategies. This post breaks down the fundamentals of traditional TAR tools and explains how and when they still best fit into today’s document review workflows.
Technology-Assisted Review (TAR) refers to the broader category of AI-powered review tools, spanning predictive coding, Continuous Active Learning (CAL), and emerging generative AI (GenAI) technologies. “Traditional” TAR, or CAL-based tools, use a combination of human review and machine learning to help legal teams identify relevant documents efficiently. Traditional TAR tools analyze patterns in reviewer decisions and apply those patterns across the larger document set to “predict” which documents are likely to be on one side of a binary classification, such as relevant or not relevant. This is different than emerging GenAI-based tools which use large language models (LLMs) to understand, classify, summarize, and analyze documents based on natural language prompts.
Traditional TAR is sometimes referred to as “predictive coding” or simply “machine learning.” While terminology can vary, the underlying concept remains the same: the technology “learns” from human input and uses that learning to prioritize and classify documents at scale.
Importantly, traditional TAR does not replace the need for human input and legal analysis. Instead, it augments human legal review by reducing the volume of documents requiring “eyes on” analysis, driving the focus toward the most relevant information first.
I will mainly focus my discussion here on traditional TAR tools, specifically CAL (or simply, Active Learning). While emerging GenAI-based TAR tools have great promise and are being incorporated into review workflows with increasing regularity, they have not (yet) replaced the use and effectiveness of traditional TAR in practical applications. In fact, the different generations of TAR tools can compliment each other.
At a high level, Active Learning workflows follow an iterative process:
1. Initial Review: Human reviewers with substantive knowledge of the matter begin to review and code documents identified as Active Learning eligible (typically those with enough text available from which the tool can learn). These decisions serve as the foundation for the Active Learning model.
2. Machine learning analysis: The system analyzes the coded documents to begin to identify patterns related to a binary classification, typically relevant or not relevant.
3. Prediction and prioritization: The Active Learning model applies those patterns across the broader document set, ranking or categorizing documents based on their predicted likelihood of relevance.
4. Ongoing refinement: As reviewers continue to review and code documents, the Active Learning model continues to learn and improve its classification or ranking. Effective quality control and validation at this stage works to ensure accuracy and defensibility of the process.
This iterative approach allows legal teams to adapt review strategies as the case evolves—while maintaining transparency and control.
Traditional TAR, or Active Learning, has long been standard tool in legal document review for several reasons:
Efficiency at scale
For large-scale document collections, “eyes on all” or linear review of hundreds of thousands, or even millions, of documents is costly and time-consuming. The effective use of traditional TAR can dramatically reduce review hours by identifying and removing low-value documents with high confidence.
Cost control
By identifying the most (and least) relevant documents sooner, traditional TAR can help legal teams manage discovery budgets more effectively.
Consistency and accuracy
Human review over larger document sets can vary in the application of substantive interpretation to coding decisions. Properly validated Active Learning models can ensure a more consistent and accurate application of coding decisions across datasets, thus reducing coding variability and improving overall review outputs.
Judicial acceptance
The application and use of traditional TAR tools have been widely accepted by courts for years. When effectively implemented and validated, Active Learning workflows are considered defensible and reasonable.
While traditional TAR can be a powerful review tool, it is not always necessary—or appropriate—for every matter. Understanding when to use Active Learning is just as important as knowing how to use it.
Active Learning is well suited for:
Active Learning may not be the best fit for:
In many matters, a hybrid approach—combining TAR with targeted eyes-on review—delivers the best results.
Despite its maturity, TAR is still surrounded by some misconceptions that can slow adoption.
“TAR replaces lawyers.”
TAR, in fact, not only requires sound legal input to be successful. It can support lawyers in developing their legal strategy by prioritizing key documents and reducing inefficient focus.
“TAR is a black box.”
Traditional AI review tools offer transparency, validation metrics, and audit trails that support defensibility.
“TAR is only for massive files.”
While the efficiencies that can be gained by utilizing Active Learning shine bright in large matters, it can certainly be a valuable tool in document collections of variable size where efficiency and cost control are priorities, and the documents themselves are conducive for use with the technology.
Technology alone does not guarantee success. The effective use and application of TAR tools require engaging in thorough data analysis, thoughtful workflow design, detailed oversight, and accurate validation measures.
This is where partnering with the right eDiscovery experts makes all the difference.
At Array, TAR tools are integrated into our client-centric, end-to-end eDiscovery consultation services whereby we define and recommend flexible, matter-specific workflow options. You choose the right balance for your specific matter, scope, and needs.
Array’s approach blends advanced technology expertise with deep industry knowledge to work with legal teams reduce risk, control costs, and maintain confidence throughout every stage of the process.
Traditional Technology-Assisted Review is long past being considered an emerging concept—it is a proven, court-accepted tool that helps legal teams manage the realities of modern eDiscovery. Understanding what TAR is, what the available tools are, and when to best use them (either independently or in conjunction with one another) empowers legal teams to make informed decisions and approach AI-enabled review with confidence.
When TAR workflows are applied thoughtfully by experienced eDiscovery professionals, legal teams can work smarter, not harder—delivering better outcomes in an increasingly data-driven legal landscape.