Ask allows you to explore your documents using natural language questions. While it’s easy to use, following this recommended workflow will help you get the most accurate and actionable results:
.jpg)
Scope – Choose how much of your dataset Ask should review to generate its answers.
Submit – Craft and submit your question to Ask.
Refine – Alter your question if you would like to improve Ask’s response.
Validate – Ensure Ask’s responses match the source documents.
Review – Decide how to incorporate your findings into your review workflow.
I. Scope
Define the range of documents Ask will search. You can choose between:
All Unculled Docs: Ask reviews every unculled document in your dataset.
Current Search: Ask limits results to a specific slice of the dataset, such as filtered search results, selected custodians, or communications within a time period.
Narrowing your dataset may be useful before asking questions (e.g. limiting documents to a specific date range or custodian).
II. Submit
Ask a question and provide additional instructions, if needed. For example:
Question: What is Enron?
Specific instructions: Organize your response in a bulleted timeline of important life events.
Though optional, adding instructions can tailor Ask’s responses to your workflow or project’s needs.
III. Refine
If Ask’s first response doesn’t provide what you’re looking for, tweak or reframe your question entirely and try again. Evaluating Ask’s responses and refining your questions can help you quickly retrieve the most useful results. Some options include:
Adjust scope – Try searching a more focused subset (e.g., emails in a date range) or expand to the full dataset.
Include duplicates – Ask normally indexes unique/pivot docs only. Adding duplicates may surface more relevant material.
Reframe the question – Phrase it in a different way to match how information might appear in documents.
IV. Validate
Just as with any research method, it’s essential to validate the answers Ask provides. AI-generated summaries are helpful starting points, but they must be checked against the actual source material to ensure accuracy and completeness.
To help you validate results, Ask provides multiple ways of comparing and exploring your dataset alongside Ask’s generated answer.
Cited documents – Clickable numbers in Ask’s answer that display exactly where the information came from, enabling quick review of the original material.
Semantic Matches – An interactive bar visualization that includes all documents with a similarity score of 30% or greater, used to filter or explore data by high, medium, or low similarity to your query. (See How Ask Operates for an explanation of how similarity score is calculated.)
High-scoring documents report – An exportable CSV that lists all high-scoring documents with their Doc IDs, scores, and relevant passages, pointing you towards other documents that may be relevant to your query.
Important
Think of validation as a critical checkpoint. Ask helps you find relevant information quickly, but reviewing the underlying documents ensures that your conclusions are accurate, complete, and actionable.
V. Review
Once you’ve validated Ask’s results, the next step is to decide what to do with them. Ask is not just about getting answers, it’s about moving your matter forward.
After reviewing the responses, you can:
Tag documents – Apply tags to relevant documents for organization, privilege review, or production prep.
Save into a search set – Group the sources you validated into a saved search so you can easily return to them later.
Export or share – If appropriate, export results or share them with your legal team for further analysis.
Ask also has a history feature that lets you revisit past queries, including seeing the search criteria used in your session.
Best practices
When using Ask, it’s important to approach the results thoughtfully. The feature is designed to help you surface insights quickly, but human review and judgment remain essential. Therefore:
Always verify Ask’s results; AI-generated responses should be confirmed against the source.
Start broad, then narrow your questions as insights emerge.
Use Ask for investigations, early case assessments, or fast fact-finding in large datasets.