Last time, we continued our “nuts and bolts” series of artificial intelligence (AI) for legal professionals with a look at several current legal use cases to which AI is being applied today that have demonstrated proven benefits.
However, generative AI and large language model (LLM) technologies are beginning to play a significant role in eDiscovery and eDiscovery software providers are rapidly adding generative AI capabilities to their solutions. In this post, we will discuss some of the emerging use cases for generative AI in eDiscovery and what the future looks like for AI – especially generative AI – in legal and some of the potential roadblocks that might impact generative AI adoption.
Seven Emerging Use Cases for Generative AI
There are numerous potential use cases for generative AI in eDiscovery, and we’ve probably only scratched the surface of what can be done with the technology. However, here are seven use cases for generative AI that are showing great potential for streamlining eDiscovery workflows:
Early Case Assessment
By analyzing the available electronically stored information (ESI) early in a case, generative AI can help legal teams assess the strengths and weaknesses of a case, potential risks, and the scope of relevant data. It can literally be as easy as loading the litigation complaint into the model and requesting documents that potentially support or refute allegations from the complaint and generating a timeline which includes those documents! This enhanced analytical ability can inform strategy decisions and potentially lead to earlier settlements or adjustments in litigation strategy.
Data Extraction and Document Summarization
Another use case for generative AI that is becoming popular is the ability to extract key facts, dates, and figures from ESI and provide summaries of evidentiary documents and even depositions, with citations to the excerpts being summarized! This capability streamlines the process of understanding the contents of documents without having to read through every document in detail.
Document Review and Classification
Generative AI models can automatically review and classify the document collection by relevance, privilege, and other case-specific requirements. Not only can the technology classify each document, but, in many cases, it can also provide a confidence level for that classification and even a short description as to the reason for the classification!
With predictive coding having been in use in eDiscovery for over a dozen years, some have questioned whether generative AI capabilities will complement predictive coding workflows or whether it will even replace them eventually. Time will tell in terms of generative AI’s ultimate role in document review – expect considerable evolution as legal teams continue to learn the best ways to apply these new tools.
Identification of Privileged Information
Generative AI models can help identify potentially privileged communications and flag them for review, which facilitates the ability to assert legal privilege and help reduce the potential of privileged and sensitive information being inadvertently disclosed.
Anomaly and Pattern Detection
Generative AI can identify unusual patterns or anomalies in data that may indicate important information or misconduct, which facilitates investigations and compliance monitoring. Examples of anomalies and patterns that generative AI can detect include variations in email exchange before or after significant events, unusual references or terminology not commonly found in the rest of the dataset (which might indicate deception), shifts in the tone or sentiment of communications over time and unusual transactions or transaction patterns.
Language Translation
For document collections involving documents in multiple languages, generative AI translation can translate documents to enable legal teams to quickly understand and categorize foreign-language documents without the need for extensive human translation services.
Question Answering and Legal Research
AI models can assist in answering legal questions and conducting legal research related to the case, helping to identify relevant laws, precedents, and legal arguments more efficiently than traditional research methods.
Four Caveats with Emerging Generative AI Use Cases
While the emerging generative AI use cases discussed above have great potential to significantly streamline eDiscovery workflows, they haven’t fully demonstrated proven benefits over an extended period of time. There are at least four caveats that will need to be addressed before we see widespread use:
Accuracy and Reliability
As we’ve seen in numerous cases (like this one we discussed in our first blog in the series), generative AI models have been known to “hallucinate” and provide inaccurate results. When it comes to document summarization and document classification – especially for complex documents – it will be important to develop a sound process for testing and verifying the results.
Interpretability and Explainability
One of the biggest impediments to the acceptance of predictive coding within litigation were concerns over the models being a “black box”, which led to parties in litigation having to negotiate the use of predictive coding and how the results would be validated. The same challenge may occur with the use of generative AI processes for document classification.
Data Privacy and Security
The use of AI in eDiscovery involves processing large volumes of sensitive information, including personally identifiable information (PII), and the use of that information in training the model. Any process for testing and verifying the results of AI models will need to include testing to minimize the risk of inadvertent disclosures of privileged, sensitive, or private information.
Costs and Accessibility
Of course, there’s also the costs associated with implementing and using these new capabilities – you didn’t think they were free, did you? For many of the new generative AI use cases, costs for using them are not fully clear, which makes determining the return on investment (ROI) for them difficult, if not impossible. Clarifying the cost picture for both implementing and using generative AI capabilities is essential for widespread adoption.
Conclusion
The application of AI to legal and eDiscovery use cases is here to stay. Legal professionals who choose to be cautiously proactive and embrace understanding AI and learning how it can be applied to support legal use cases will be ahead of luddites who resist the use of emerging technology and strive to keep things “the way they’ve always been”.
Leave A Comment