Assist (aiR Assist)

Assist is a conversational search tool integrated within a Relativity aiR workspace, designed to empower legal teams to interact with their data using natural language. By leveraging advanced AI, Assist helps to surface potential insights, reveal possible connections, and uncover themes more efficiently. This can enhance the process they use to analyze and interpret legal data more effectively, potentially leading to quicker understanding, better decisions, and defensible workflows when validated by users.

It works by searching the extracted text of indexed documents. Users can create up to five indexes per workspace, each supporting up to 300,000 documents. When a query is submitted, Assist identifies the documents deemed most relevant and employs a large language model (LLM) to generate answers grounded by citations tied back to the documents.

For additional information on AI and aiR products, see AI products and features in Relativity aiR.

Refer to Frequently asked questions for helpful answers to common questions about using Assist.

Assist availability

Assist is available to organizations with an Case Strategy contract or aiR Integrated contract. It does not require installation in Relativity aiR.

Contract type	Availability
Case Strategy contract	Assist is included in Case Strategy at no additional cost and does not draw down on document thresholds. It is available in review workspaces where Case Strategy is installed and at least one Fact Extraction or Transcript Summary job has completed. Customers can also create a Case Home index. For more information about completing these jobs, refer to Case Strategy (aiR for Case Strategy) documentation. If an Case Strategy workspace does not meet the required conditions (Case Strategy is installed and a Fact Extraction or Transcript Summary job has completed), users can open Assist, but they cannot use it until one of those jobs runs.
aiR Integrated contract	Assist is included in Integrated contracts at no additional cost and does not draw down on document thresholds. It Assist is available in both repository and review workspaces. The Case Strategy installation and job-completion requirements do not apply. Customers can also build a Case Home index based on Case Strategy documents. For more information about completing these jobs, refer to Case Strategy (aiR for Case Strategy) documentation.

To use Assist, the appropriate permissions must be configured for each group in each workspace, regardless of contract type. See Permissions for more information. By default, System Administrators are granted access to all Assist permissions across all workspaces.

The Assist icon appears at the top of the sidebar.

Index and document limits

Below are the index and document limits for Assist.

Type	Limit
Index	Each index can contain up to 300,000 documents. A maximum of five (5) built indexes can exist at one time per workspace. This can be from a combination of the Case Home document set (created in Case Strategy) and public saved searches. See Working with indexes for details on adding and deleting indexes.
Document	Individual documents must be up to 5 MB of extracted text; larger files are excluded during indexing. Only documents with extracted text are indexed. The text must be stored in Data Grid (not SQL). Documents that do not have extracted text are automatically excluded from the index.

Basic getting started steps

Some initial configuration must be completed before anyone can begin using Assist. Below are some basic workflow steps to getting started with and using Assist.

Task	Description
1. Configure user group permissions	User(s) granted Access Management permission (typically a System Admin) assigns permissions to users, such as Index Management and Prompting. See Permissions for details.
2. Map the metadata	If using metadata, user(s) granted Index Management permission (typically a System Admin) maps the metadata fields. See Metadata mapping for details.
3. Create indexes	User(s) granted Index Management permission creates 1-5 indexes (document sets) based off of public save searches that will be used by those with Prompting permission to generate responses to their prompts. Working with indexes.
4. Run a Fact Extraction or Transcript Summary job (only for Case Strategy contracts)	If your organization has an Case Strategy contract, at least one Fact Extraction or Transcript Summary job must be completed. See Case Strategy (aiR for Case Strategy) documentation for details.
5. Ask questions/prompts	Users granted Prompting permission enter questions/prompts against a selected index to generate responses and document citations. See Prompting for details. Also refer to Best practices for guidance on creating prompts and Using Conversation Manager for managing conversations.
6. Review and verify responses	Users granted Prompting permissions review and verify the responses in the Conversations panel, Document list, and Viewer. See Navigating the responses for details. Also refer to Best practices for guidance on indexing.

How Assist finds answers

Assist operates using a Retrieval-Augmented Generation (RAG) process to deliver grounded, evidence-based responses. This approach combines document retrieval with LLM generation to help support accuracy, transparency, and contextual relevance.

1. Asking a question (question step)—The user asks a question aimed at the index (document set) they selected to use.
2. AI analyzes the question and plans the search (analyzing step)—The question is read by the LLM, which decides whether the question is an appropriate one to answer. If so, it formulates queries to run against the index to find documents.
3. Searching relevant documents (search step)—Assist performs one or more searches to identify the most relevant content within the index of document chunks.
4. Generating the answer with links to cited documents (generation step)—The selected passages, along with the original question and system prompt, are passed to the LLM. The model uses this retrieved context to generate a response intended to be coherent, concise, and supported by retrieved content, citations, and references to the original sources.
5. Links to cited documents (answers step)—The user reviews and verifies citations in linked documents.
   

Understanding Assist responses

Assist helps identify and summarize potentially relevant information across large document sets using natural language interaction. Built on a Retrieval-Augmented Generation (RAG) architecture, it retrieves and analyzes the documents most likely to be relevant, then generates a citation-supported response based on that content.

It returns contextually relevant and evidence-based information rather than performing exhaustive or “find everything” searches. Because it does not review each document individually, some keyword or topic matches may not be included in the response.

The RAG process works best when key evidence is found in a few focused documents. Results are less accurate if answers depend on scattered or unclear information.

It is also important to note that document security may differ between users. As a result, two users could ask the exact same question and receive different answers, because the LLM interaction used for each response only knows about documents that the requesting user is authorized to access. Assist does not retain knowledge from other users' conversations, and every interaction is evaluated against the requesting user's document security permissions at the time of the prompt.

Language support

The LLM used by Assist has been evaluated for 83 languages. Although Assist has primarily been tested on English-language documents, it is designed to support non-English datasets. For more information, see Language support in aiR products.

Unlike other aiR products, Assist system consists of two components:

• Retrieval—the LLM generates search queries to find relevant documents in the index.
• Generation—the LLM summarizes the retrieved documents, picks relevant ones, and synthesizes them into an answer.

Because of this two-step design, there are special considerations when working with non-English datasets. If you use Assist with non-English datasets, we recommend the following:

• When possible, write your question in the same language as the documents being queried—ideally one you are fluent in. If that is not possible, write your question in English.
• By default, Assist searches for documents in English. If your documents are in another language and you are writing your question in English, tell it which language to search in by adding a sentence such as: "Search in <Language>." This helps the retrieval step generate queries in the specified language. While effective in most cases, the LLM may not always apply the requested language consistently.
• Extracted text and citations remain in the language of the source document — you do not need to translate them yourself before reviewing.
• By default, answers are generated in English. To request an answer in another language, add a sentence to your question, such as "Write answer in <Language>." This approach is effective in most cases, but the LLM may not always generate answer consistently in the requested language.

You can inspect which searches were attempted using the Search Completed summary in the response panel (see Navigating the responses). Use this to iterate on your question and adjust language coverage as needed.

Common use cases

Here are some example questions targeting a few common use cases for Assist:

Use case	Common category	Example question
Early Case Insight	Finding potentially key/important documents	Can you find me documents that discuss potential gifts or incentives?
	Finding documents by theme	Are there any documents mentioning fraudulent behavior of John Doe?
	Understanding actors and roles	Who was involved in discussions about offering gifts?
Case Strategy Development	Identifying a series of events	Create a high-level timeline for events that took place before the start of Project Artemis.
	Understanding communications and relationships between actors	Who communicated with whom about the contract terms?
Deposition/Trial Preparation	Suggesting exhibits based on key criteria	List documents to use as exhibits based on [key document criteria].
	Confirming conversations or actions took place	Did John Maxwell send an email about the compliance policy?

Additional scenarios include but are not limited to:

• Evidence identification—query an issue to quickly surface relevant, citation-backed documents.
• Communication analysis—map interactions and relationships between custodians and actors.
• Production deficiency review—test incoming productions against RFP requests to identify gaps.
• Entity and role understanding—clarify who's who and their responsibilities across the document set.
• Sample set identification—find candidate documents to use in the relevant sample set for an aiR Review project.

Auditing user activity

You can monitor Assist user activity in the Audit application in your Relativity instance. The Audit table records the events listed below. Because Assist audit events are not RDO objects, they use an Audit Object UUID in the Audit table instead of an Object Artifact ID. For more information on using the Audit application, see Audit.