Everlaw AI Assistant: Can this AI Help Cross Your Project Over the Finish Line?

What is Everlaw AI Assistant?

On 25 April 2024, Everlaw announced that its generative AI offering known as Everlaw AI Assistant (simply referred to as “Everlaw AI”) was available for all its customers to use in Beta mode. This is nearly a year after it was rolled out to a select number of clients only.

Everlaw AI is a cloud-based platform that utilises the Large Language Models (LLMs) developed by OpenAI (i.e. the creators of ChatGPT). Among other things, it provides insights into cases by creating document summaries, assists with drafting legal documents and provides support on the first-level review process within an Everlaw workspace. These functions are similar to other tools previously posted on this blog.

The Key Features

There are 5 features available with Everlaw AI:

1 - Coding Suggestions

Everlaw AI can provide coding (or “tagging”) suggestions which can help legal teams to categorise and review documents more efficiently. This includes Relevance, Privilege, Issues and other options. Users can then decide whether the tags should be applied to the suggested documents or not. For it to work, users will need to input the case description and coding criteria (also known as “prompts”) which make it applicable to the tags. The coding criteria can also be amended at any stage.

2 - Document Summarisation

This feature generates concise summaries of documents for case teams to review instead of having to review an entire document. This in turn saves valuable time and effort as large documents with mostly irrelevant content can be reviewed quickly for relevance and other issues. For instance, if a witness statement arrives moments before a scheduled meeting with a client, this can be helpful to quickly understand the overall picture of the information received and in time before your meeting starts.

3 - Entity Extraction

This option enables users to identify information about people and organisations within documents in order to review the most relevant information first. Identifying the entities within your dataset is a good method to quickly filter your documents for review, especially when you know what you’re looking for.

4 - Sentiment Analysis

Users can identify relevant documents based on the “sentiment” (positive or negative) tones within the text of a document. This is beneficial for case teams who wish to uncover documents which contain emotions that lead up to the significant turning point on a case. For example, identifying the points in a contract dispute where negative sentiments arise can help illustrate the sequence of events and breakdown in communication. This could affect the entire legal outcome.

5 - Open-Ended Queries

Perhaps a new area for many legal professionals in document review, this feature enables users to ask questions about the information contained within their documents in Everlaw. This could be whether a certain person communicated with another key player and how often. The possibilities with this functionality are endless and it takes a creative mindset to grasp its full potential in order to discover the truth behind the facts.

It’s always recommended that human judgment is used and responses generated by AI are checked by users to ensure the answers are sufficiently backed up by evidence.

Everlaw’s Principles on GenAI

Everlaw has five key principles on the use of generative AI on its website.

It starts with giving its users control which involves giving Project Admins the ability to toggle its AI functionalities at the project level. In addition, the AI features are designed to give confidence to users, for example, by producing justifications for its decisions based on the evidence contained within an Everlaw workspace. Everlaw also aims to maintain transparency, privacy and security when they use third-party tools. Overall, Everlaw recommends exercising caution and states that human validation is still crucial for some AI-generated content regarding the facts of a case due to the nature of LLMs and their ability to make mistakes.

The information contained within an Everlaw workspace is SSL encrypted which means that the data inputted into the AI model is secure and not used to train other OpenAI models. Everlaw follows standard data privacy and compliance requirements through the GDPR and other legislation.

Limitations

Like many GenAI tools, Everlaw AI has its limitations. Here are some key points to consider when using Everlaw AI:

• Limited to Textual Data: It won’t perform well on documents which do not contain any text, such as photographs, and even data contained in tables as the formatting can either be lost or “misunderstood” which leads to inaccurate output. Therefore, it is important to ensure that you check the results it produces against the documents it references.

• Accuracy of Generated Text: The LLMs used by Everlaw AI are trained to produce fluid text (i.e. text which changes continuously) and are not necessarily accurate responses. They may produce false or misleading statements. Therefore, it’s important to exercise judgment and expertise to validate responses produced by generative AI.

• Requires Human Validation: Particularly when creating factual claims about people or events, or work product intended to be shared with others.

These points highlight the need for legal teams to appreciate that AI tools are not a substitute for human expertise and judgment. They should be used in conjunction with the skills and knowledge of qualified lawyers.

Conclusion

Everlaw AI contains a range of features that can streamline eDisclosure workflows.

The implementation of Everlaw AI is significant as it moves away from traditional human analysis and electronic document review. By automating many of the time-consuming tasks associated with the review and trial preparation, it allows legal teams to focus on more important elements of their work. For example, instead of reviewing 100s of irrelevant documents, legal teams can find the 1 key document which is relevant to the issue in question and move on to other tasks like preparing an email with legal advice to a client. This is a positive step forward and clearly highlights there is no intent for AI to replace a legal professional’s job.

Moreover, the ability of Everlaw AI to provide insights into the information contained within data can help legal teams better understand the background of their cases faster and mean they are more likely to meet their deadlines. This is ultimately going to lead to more successful legal outcomes.