What is the maximum case size that Ask can support effectively?
Ask is designed to support large-scale cases containing millions of documents within our SaaS environment.
What is the maximum document text size that Ask can search?
Ask indexes only the first ~350K text of a document. If a document has more than ~350K text, the content beyond this threshold is excluded from indexing and the “Partial Ask Indexed” auto-tag will be checked for that document.
What content does Ask analyze when generating responses?
Ask evaluates up to 20 references when generating a response, selecting those deemed most relevant to the submitted question. A subset of these references is then used for citations in the final answer.
How many documents does Ask return?
Ask returns up to 100 documents that best match the submitted question. From this set, it analyzes up to 20 references identified as the most relevant for answer generation. A subset of these references is then selected for citations in the final response.
What are the best practices for crafting effective questions for Ask?
Be explicit.
The more specific you are with the queries you pose to Ask, the more likely you will be to get helpful responses. It helps to avoid ambiguity.
Avoid: “Who oversaw handling of reporting for the company?”
Try: “Who oversaw handling of financial reporting, such as income statements and balance sheets, for Enron?”
Narrow in scope and can have a clear answer.
Narrow questions improve focus, reduce ambiguity, and yield precise, actionable answers.
Avoid: “Tell me everything about how the company handled its finances.”
Try: “Were there any documents from finance raising concerns about improper financial reporting?”
Pose one single question at a time.
It is better to use queries that pose single questions rather than queries that pose compound ones.
If the question you have in mind has two parts, ask each part as a separate query.
What types of questions are less effective when using Ask?
Avoid requests to identify “all” documents or “all” items in the case. E.g.:
“Give me all finance reports.”
“Find all insurance policies in 2025.”
“Retrieve all medical records in the case.”
Avoid questions requiring a broad analysis across multiple documents.
“Was there fraud at Enron?”
“Is Andy Fastow guilty?”
How should the “Additional Instructions” feature be used?
Additional Instructions enable users to guide Ask in performing further actions on the retrieved content, such as analysis, transformation, and formatting. In general, action and formatting directives should be kept separate from the main question. This approach improves clarity and helps ensure more accurate results.
For example, instead of combining everything into a single query: “Are there specific concerns regarding the accuracy, transparency, or integrity of Enron’s financial reporting practices? Return results and summarize it in 200 words.” It’s preferable to structure the request as follows:
Question:
“Are there specific concerns regarding the accuracy, transparency, or integrity of Enron’s financial reporting practices?”
Additional Instructions:
“Return the results and provide a 200-word summary.”
Does Ask support questions submitted in languages other than English?
Yes, you can find a list of languages in our Supported Languages in Ask article.
Does Ask retain a history of previously submitted questions and answers?
Yes, Ask History helps you keep track of your past questions. See How to Use Ask for more information.
What is the difference between “Documents” and “References”?
References are excerpts derived from the source document. During data ingestion, Ask segments each document into multiple references to optimize performance.
Can Ask analyze multiple documents simultaneously?
Ask performs its analysis based solely on the top 20 retrieved references. These references could be from multiple documents, but it does not performs analysis beyond the top 20 references.
How can users validate the answers provided by Ask?
Answers in Ask can be validated using the citation feature. See Ask Citations.
Are questions I ask visible to other users who have access to the same project?
No, Ask History in-app is only visible to the user themselves. However, account or project admins have the ability to download a report containing all sessions from all users in the project.
Can I control which users have access to Ask AI button?
Yes, you can control user access to Ask using user access levels. Project-level access to Ask is available to Basic Users and above. Users with the Reviewer or Limited Reviewer roles currently do not have access. More information can be found in User Access Levels Overview.
What’s the method Ask uses to retrieve documents before analyzing?
Ask retrieves a list of the of most semantically similar documents related to your question. For more details, see How Ask operates.