Enroll and Get Started with Deep Dive

Deep Dive is Everlaw's generative AI powered question and answer tool, which lets you and your team ask natural language questions to get citation-backed answers grounded in the contents of your documents.

This article is for admins responsible for enrolling databases and projects into Deep Dive. Use it to understand how Deep Dive enrollment works and some enrollment best practices that help you make the most of the tool, as well as for detailed instructions on how to enroll databases/projects in Deep Dive.

How does enrollment work?

Any organization with a contract that includes Everlaw AI can enroll in Deep Dive. To enroll a project or database into Deep Dive, you must be an Organization Admin or Everlaw AI Admin. See the enrollment section for detailed instructions for how to enroll.

Deep Dive enrollment is done per database. When you choose a database to enroll, you can select either:

• The entire database, which includes all projects (both complete and partial) within it
• One or more partial projects. This can be a good option for those seeking to ask questions of a targeted subset of documents. See the Best practices section below for more details on leveraging partial project enrollment to control Deep Dive costs.

Here are some important things to be aware of when you enroll:

• Deep Dive cost: At the time of enrollment, there is an estimated max GB cost for the data being enrolled. The enrollment dialog displays the projected cost in credits, based on the total amount of data in the project(s) being enrolled, and is rounded up to the next whole credit. 

  Note

  If you are enrolling multiple partial projects, the final cost can be lower because documents that exist in more than one partial project within a database are only billed once. See our Manage and Track Everlaw AI Credit and Feature use article for details on how to track Deep Dive credit costs.

• Deep Dive credit usage is unlimited: The enrollment will proceed and the credits will be used regardless of whether your organization has set a Batch Review Assistant credit limit. This ensures that all documents in an enrolled project can be queried when answering questions. If there are no prepaid credits remaining, on-demand credits will be used. See our Everlaw AI credit tracking article to learn more about tracking credit usage.
• Cost for future data additions: Any future data added to enrolled projects will incur the per GB cost at the time it is added. This includes new uploads, documents promoted from ECA into an enrolled review project, or unenrolled data that is later added to an enrolled partial project (e.g. documents uploaded to a complete project that is not enrolled in Deep Dive and later added to an enrolled partial project). There is no option to exclude data from being enrolled into Deep Dive when it's added to an enrolled project.
  Productions you create on enrolled projects are enrolled into Deep Dive, but do not incur credit usage for their enrollment. 
• The credits used for Deep Dive enrollment are a one time cost for the lifetime of the document: Unlike data hosting fees on Everlaw, which are monthly, the credit cost for Deep Dive enrollment is a one time charge for the lifetime of the document. This means:
  • For each data addition (at the time of enrollment and later uploads), you will see a one time credit usage. Credit usage events appear upon completion of each enrollment, and each enrollment cost is rounded up to the next whole credit. Some tasks might be tracked in one more than one row (more than one event). 
    Productions you create on Everlaw do not incur credit costs.
  • Once data has been enrolled, there is no subsequent or recurring cost for Deep Dive to access that data.
  • If you enroll an entire database and create multiple separate partial projects within it, you are not charged again for documents that are added to multiple projects. Once a document has been enrolled, it can be accessed by Deep Dive in any enrolled project within the database at no additional cost.

• You cannot unenroll: Once a database/project is enrolled in Deep Dive, you cannot unenroll it. You can reach out to support@everlaw.com to have the Everlaw team manually unenroll from Deep Dive, but this will also delete the entire question and answer history from Deep Dive. We do not recommend unenrolling. 
• You cannot update enrollment: Once you complete the steps to enroll, you will not be able to later enroll new partial projects within a database that already has enrolled partial projects, or “downgrade” a complete project enrollment to partial project enrollment.
• For users transitioning from the beta program: Everlaw AI must be enabled from the Project Management > Project Settings > Everlaw AI page to use Deep Dive on any already enrolled projects. Even for an enrolled project, Deep Dive will not be available if Everlaw AI is not enabled at the project level. 
• Everlaw AI is enabled by default when a project is newly enrolled: When you complete enrollment on the Organization Everlaw AI page, Everlaw AI is automatically enabled on any enrolled projects. This means that all single action (no cost) AI features are available to users on the project with the proper permissions. Organization Admins and Project Admins can disable any of these features from the Project Settings page.

• The initial enrollment must complete before you can start asking questions in Deep Dive: You can continue asking questions during subsequent data enrollment, however answer quality may vary during this time. The Deep Dive interface displays a status box to indicate that document enrollment is taking place. 
  

What data is eligible for Deep Dive? 

Deep Dive uses your documents' text to generate its responses. A simple way to think of it is that, if content is present in your document's text file, it is accessible to Deep Dive. 

If you run productions on an enrolled project, Everlaw-produced documents are automatically ingested into Deep Dive. Deep Dive waives the enrollment cost for these produced documents.

Metadata

Deep Dive uses the file name and file type of the retrieved documents to determine their relevance. Deep Dive doesn’t currently consider additional metadata, such as primary date or custodian, in its relevance determinations.

To ask Deep Dive questions based on specific metadata values, build a search of documents restricted to those values, then use Deep Dive’s Ask a question button on the results table to query just the documents meeting those results.

Non-English text

Deep Dive works with text in many languages, though the embedding model and LLMs behind Deep Dive don't have equal training and context for all languages. Typically, a response will be returned in the dominant language of the text used to generate the response. For example, if the user asks a question in Spanish, and most of the documents used to generate a response are in Spanish, then the response will most likely be in Spanish, too.

Best practices for enrolling in Deep Dive

Because all documents with text added to any enrolled project incur a credit cost, it's best to have an understanding of your total data quantities, both current and projected, before enrollment. This can help you manage your costs and avoid unexpected charges.

One strategy is to create an empty partial project within a database, name the project to indicate that it is enrolled in Deep Dive, and enroll that project. Because uploaded documents are not automatically added to partial projects, your team has complete control over which documents are added to it, and therefore which documents incur ingestion costs. 

Once you have made the empty project and enrolled it, you add any documents that you want to enroll in Deep Dive into that partial project. When you do, they will be enrolled and incur the enrollment cost. 

While this strategy gives you granular control over Deep Dive costs, it also requires more manual work and attention. As documents are uploaded, an admin needs to make sure all the documents your team wants to ask Deep Dive about are added to the enrolled partial project. 

It's also important to understand which documents are not in Deep Dive and to make sure that they receive the review and attention that is required. We recommend that you make binders on the complete (unenrolled) project to keep track of which documents are being worked on in the enrolled project and which are available solely in the complete, unenrolled project.

If you need to produce documents spanning those in both the complete and partial projects, you should copy the review work from the partial project onto the copies of the documents in the complete project, and run the production from the complete project.

Enroll in Deep Dive

Requirements: Deep Dive enrollment is controlled at the organization level. To enroll databases or projects, you must be an Organization Admin or Everlaw AI Admin.

To enroll a database or project:

1. Go to Organization Home > Everlaw AI > Credit management. 
2. In the Database credit usage table, find the Deep Dive enrollment column, and select Enroll for the database you want to enroll.
   This opens the Enroll dialog.
   You decide in the next step whether to enroll the entire database or just select projects.
   
3. Select either Entire database or Partial projects. 
4. [Optional] If you selected Partial projects, use the checkboxes to select which partial project(s) to enroll.
   

5. Select Next to move onto the confirmation step.
6. On the confirmation step, you confirm two important aspects of Deep Dive enrollment:
   1. All current documents in the selected projects, excluding documents created in a production, will be enrolled and incur credit costs. Here are some additional details about enrollment credits:
      • This checkbox includes the highest possible credit cost for enrollment of the selected projects
      • A document is only charged once, regardless of whether or not it is in multiple partial projects being enrolled
      • Credits are rounded up to the nearest whole credit
   2. Future document uploads in those projects are automatically enrolled and will incur credit costs. The documents from future productions that you create in enrolled projects will be enrolled but will not incur a credit cost.
7. Select Confirm & Enroll. This starts the enrollment process. This process prepares documents for Deep Dive. The time to finish enrollment depends on the amount of data and total file count in the enrolled projects/databases. 
   If the estimated enrollment time is excessively long, you will see a banner stating so. To enroll data from this database, you can add a subset to a partial project and enroll the partial project. As the technology behind Deep Dive improves, we expect to ease limits on enrollment.