AI-related discourse has increased in the legal industry over the past two years, with a particular emphasis on eDiscovery. The marketing of AI is constantly evolving, frequently resulting in exaggerated promises or mixed messages that conflate human automation with genuine AI capabilities. There is a tendency to disregard these developments and adhere to conventional workflows amid the hype. Conversely, legal professionals are ethically obligated to comprehend and capitalize on contemporary AI innovations, including large language models (LLMs), in order to optimize eDiscovery procedures.
When properly trained and implemented, the appropriate AI can offer substantial advantages, enabling law firms to provide greater value to their clients and preserve a competitive advantage. The critical concern that arises is whether this AI technology can actually enhance my legal practice.
The Evaluation of AI’s Effectiveness in eDiscovery
Prior to incorporating AI into eDiscovery workflows, it is imperative to assess its efficacy—as in, its ability to provide the anticipated advantages. Understanding the technology that underpins AI is essential, as not all AI that is promoted as a solution is of the same quality. To evaluate efficacy, legal professionals should evaluate metrics that have a direct impact on their practice, including quality, efficiency, and return on investment (ROI).
In eDiscovery, two primary categories of AI are generating significant waves: predictive AI and generative AI. These AIs are utilizing LLMs. In their own unique methods, each serves distinct functions and influences quality, speed, and ROI.
Assessing the Effectiveness of Predictive AI
Predictive AI in eDiscovery is concerned with determining the probability that a document will be classified as responsive or privileged. The process is typically comprised of the following:
- The examination and coding of a representative sample of documents by attorneys.
- The AI model is trained on specific criteria using this coded set.
- The trained AI classifier is permitted to assess the remaining documents, designating a probability score based on the likelihood.
High quality
The performance of predictive AI is evaluated by its ability to accurately identify pertinent documents. High-quality artificial intelligence (AI) has the capacity to identify subtleties in documents that may be disregarded by human evaluators. For example, a predictive AI classifier identified 1,600 privileged documents that were overlooked by conventional search terms, thereby safeguarding the firm’s reputation and preventing inadvertent disclosures.
Efficacy
Predictive AI prioritizes document evaluation, thereby increasing speed. The documents that are most likely to be relevant are the first to be reviewed by certain firms, who employ AI-generated percentages. By substantially reducing the time spent reviewing documents that are unlikely to be of interest, this method streamlines the review process. For instance, a company succeeded in eliminating 200,000 documents from review by employing AI to prioritize and cull documents, thereby significantly reducing the time spent on routine tasks.
Earnings
Return on investment (ROI) is readily apparent when AI enhances both quality and speed, thereby providing clients with superior service and a more efficient utilization of resources. In a single instance, a law firm reduced costs by over $1 million and saved 8,000 attorney hours during a privilege review by implementing predictive AI. This resulted in a significant increase in client satisfaction and cost savings.
Gauging the Effectiveness of Generative AI
Through the generation of text or summaries in response to queries, generative AI (gen AI) provides capabilities that surpass predictive AI. It is extremely dependent on the specific use case, and its effectiveness in eDiscovery can vary. Alternatively, generative AI has the capacity to generate concise summaries of documents or respond to inquiries. Across various eDiscovery use cases, its efficacy and applications for outside counsel vary significantly.
An example of AI’s performance being linked to a training period is the use of gen AI to compose privilege log content, which increases its significance in more substantive matters. The AI can generate tens of thousands of accurate privilege entries in a day after training on a few hundred privilege records by expert analysts. The purpose of this training is to incorporate the perspective and feedback of corporate and outside counsel on privilege into the model.
Quality and Validation
Quality of generation AI outputs is contingent upon validation and control. If not properly managed, genAI has the potential to produce content that is highly accurate, but it can also deviate from the truth. For optimal performance, AI analytics professionals should supervise the AI’s outputs, ensuring that they are consistent with legal standards and regulations.
One example is the utilization of genAI to create privilege records, which results in consistent, high-quality output that can be reviewed and modified by attorneys. GenAI’s capacity to improve the quality of legal documentation was demonstrated in a test in which AI-generated privilege log lines were rated higher than those produced by novice attorneys.
Accuracy
Generative AI has the potential to significantly reduce the time required to generate initial documents. Despite the fact that AI-generated content should always be reviewed for accuracy, the initial drafting phase is expedited, enabling attorneys to concentrate on refining rather than creating from scratch. Faster turnaround times and more efficient workflows are the result. Attorneys ought to address AI-generated content as an initial manuscript, as is widely recognized. In the same way that you would not submit a draft brief written by a summer associate without reviewing and fact-checking, you would not submit genAI content without reviewing.
On the other hand, AI generates content at a significantly quicker pace than even the most adept attorney. In addition, the process of reviewing and editing a draft is significantly more efficient than the process of writing from inception (provided that the content is accurate).
Earnings
The calculation of ROI for genAI can be simple when it is applied to specified tasks. GenAI can be more cost-effective than contract reviewers or junior attorneys when it comes to drafting privilege records, for instance. Companies can enhance their overall value proposition by offering competitive pricing and attracting recurrent business as a result of this cost efficiency. Providing genAI with recognition for driving direct value for a law firm can be challenging due to the numerous use cases required. If you employ genAI as a conversational search engine or case-strategy collaborator, how do you determine its return on investment in terms of dollars and cents?
However, it is straightforward to monitor the financial ROI when dealing with specific workflow duties, such as privilege logs: How much does your organization allocate to privilege logs with generation AI in comparison to without? Certain generation AI has been discovered by numerous firms to be more cost-effective than utilizing junior attorneys or contract evaluators for the initial draft. This has allowed the firms to offer more competitive client billing, which in turn has resulted in a higher volume of repeat business.
The Identification of the Appropriate AI for Your Organization
In order to ascertain the most effective areas for AI, law firms must identify the areas within their eDiscovery processes that can be automated. The query is not merely “How can AI enhance my work?” but also “In which areas can AI provide the most value?”
Realistic expectations for quality, speed, and ROI are established by identifying the appropriate form of AI and identifying the appropriate use cases. In addition, it assists in the establishment of benchmarks to evaluate the effectiveness of AI integration.
Importance of Ethical Considerations
With the increasing integration of AI into legal practice, ethical considerations are becoming increasingly important. It is essential for attorneys to remain informed about technological advancements and comprehend the risks and benefits of AI use. Competence in AI is not solely a technological issue; it is also an ethical one, as ethical violations may result from the failure to verify the accuracy of AI-generated content.
Trust and Precision
AI tools are only as reliable as the data and parameters on which they are trained, despite the fact that they can substantially reduce the risk of human error. Legal organizations must perform exhaustive due diligence to comprehend the datasets that are employed to train their AI tools, thereby guaranteeing that these tools correspond to the rigorous standards necessary for legal practice.
High levels of accuracy are achieved when genAI tools are appropriately guided and validated. To preserve the integrity of the legal profession, it is imperative that firms guarantee that these instruments operate in accordance with rigorous ethical standards.
So, should you be utilizing artificial intelligence in electronic discovery?
The landscape of eDiscovery is being revolutionized by AI, which provides unparalleled efficiency and accuracy. Integration of AI tools, such as predictive and generative AI, can result in substantial time and cost reductions for law firms, while simultaneously enhancing the quality of legal services.
By focusing on ethical practices, assessing the efficacy of AI tools, and asking the right questions, law firms can leverage the power of AI to deliver exceptional value to their clients, remain competitive, and enhance their eDiscovery processes. The significance of AI in the legal profession will only increase as it continues to develop, necessitating that legal professionals innovate and adapt to these transformative technologies.
