Legal tasks often require searches across large sets of data, like litigation and e-discovery records, or internal banks of briefs or contracts. CoCounsel uses search queries to quickly identify responsive information.

Select a database and enter the search query in plain language. CoCounsel then searches the database and returns a list of relevant files and provides links to the full text and specific pages that answer your question. How do I create and maintain CoCounsel Core databases?

1. Getting started

At the bottom of the chat, select "⊕ Upload" and "Search an entire database." A list of your databases will appear.

2. Select a database

A list of your databases will appear. A database is in this list because either:

Use the search box and filters to find and select a database. Then click "Select database."

3. Enter the query

CoCounsel confirms the selected database. Enter the query into the chat, then press Enter or send the message.

4. CoCounsel interprets your request

Make edits, cancel, or click "Submit request."

5. CoCounsel searches the database for responsive information

A bar shows CoCounsel's progress. Click "Email me when complete" to be notified when your results are ready. You may also cancel your request or click "View results" to watch CoCounsel work.

6. Navigate and verify the results

When the results are ready, click "View results."

The results confirm the Request, provide a narrative Response, and list the most relevant supporting Documents.

Each document has a collapsible/expandable heading with a link to the full text of the document and quoted highlights of relevant sections of the document.

Click document titles for the full text, or click a pincite to open the document to that page. The document opens in a new tab in the web browser.

7. Use the results

Download or copy the results.

Select documents for follow-up.

Checkboxes will appear next to each document. Check individual boxes or check "Select all." Click "Submit" or "Cancel."

The selected documents are added to the chat for follow-ups, like a summary of each document.

CoCounsel understands nuance, humor, and implied references. This means CoCounsel finds information and references that traditional search engines cannot. The best requests are specific, clear, precise, and concise.
​

Example prompts

Searching a database of Enron emails

Query: Did Andrew Fastow know about unethical accounting practices?

This prompt shows how CoCounsel uses the context of your request to find responsive documents. Instead of asking for a list of all emails by Andrew Fastow, this prompt asks CoCounsel to search the database for emails with content indicating that Andrew Fastow had knowledge of wrongdoing. CoCounsel is essentially “reading” relevant documents in the database for context, not just a certain keyword.

Searching an umbrella patent database

Query: Does a patent exist for an anti-inversion umbrella with a removable cuff?

This search illustrates a prompt that goes beyond asking for a list of documents. Patents can be difficult to read and understand, but CoCounsel can understand and search for the context of your request to find responsive documents, even when the dataset includes complex, technical content.

Searching a database of estate planning documents

Query: What language do I need to include in a pet trust?

This prompt shows how you can create and use databases of estate planning documents, like wills and trusts, to find provisions in prior documents that can be used in current drafting projects. This prompt also makes use of CoCounsel’s ability to understand context.

This skill does not conduct a word-by-word review of all documents.

Search a Database will run searches based on your question and summarize the results that are relevant. It will not review documents that are not returned by its searches.

Avoid asking questions that require CoCounsel to perform a mathematical calculation.

Large language models can struggle with math. In addition, because this skill only provides summaries based on search results, it should not be used to answer questions that require a mathematical analysis of your entire database.

Query: ~1,000 characters

The Response will include the most relevant results. While there will not always be this many relevant results, the Response may include: