Skip to main content

Create a table

Show how to create a new analysis table and configure its initial structure.

Written by Caleb Harris

When to use this workflow

Create a table when you already know the analysis you want to perform and want to build the structure yourself instead of starting from a saved template. This is the right option when:

  • you want to name and scope a new analysis from scratch

  • you need custom columns for a specific matter

  • you want to add your own sources before running anything

  • you expect to refine the prompt design as you work

If you want a prebuilt structure, use the template-based creation path from the Tables page instead.

Start from the Tables page

Open the Tables page for the project, then use the New Table menu. Choose New Table to open the blank-table creation modal.

The list page also lets you search existing tables, so it is worth checking whether the project already has a reusable table before you create a new one.

Create the table shell

In the Create New Table modal:

  1. Enter a required value in Table Name.

  2. Optionally add a description that explains the purpose of the table.

  3. Select Create Table.

After the table is created, &AI takes you directly into the new table detail page.

Add the first sources

A newly created table is just the container. To make it useful, add rows through Add Sources.

The dialog lets you select:

  • processed project documents

  • project products

You can add both in the same action. Existing rows are excluded from the picker so you do not accidentally add the same source twice. If the project does not yet contain the document or product you need, the modal links you back to the project sources area so you can add those items first.

Add the first columns

Use Add Column to define the questions or searches that the table should run for each row. The column modal supports three setup patterns:

Create a new prompt

If your workspace allows new prompt creation, the Create New Prompt tab lets you define a brand-new prompt for the column. This is the most flexible option when your question is unique to the current project.

From that tab you can configure:

  • the prompt content

  • the output type

  • output constraints

  • the mode selector for Lite, Regular, or Max

  • whether web search is enabled

  • whether deep research is enabled

If your prompt references other columns for context, the modal can carry those column references into the execution configuration.

Select from the prompt library

Use Select from Library when a saved prompt already exists and fits the analysis you need. This is helpful when your team wants consistent columns across matters.

Use a search-backed column

Use the Search tab to set up a search execution instead of a direct prompt. The modal currently supports:

  • prior art search

  • product search

You can set how many results to return and add optional context to steer query generation.

For product-search columns, the modal also supports Minimum Annual Product Sales, Public companies only, and Include standards so you can narrow the search before you run it.

When Max Mode is available, turn it on if you want reranked results with richer reasoning and attribution in the returned matches.

Adjust the setup to fit your workspace

If your workspace does not allow creating new prompts, use saved prompts from the library or search-backed columns instead.

Prepare the table for analysis

Before you run anything, make sure the table has:

  • enough rows to compare meaningful sources

  • columns that clearly ask for the output you need

  • output types that match the review task

  • search, web, or deep-research settings only where they are actually needed

After creation

Once the table has sources and columns, use Run All or Resume All to begin token-consuming analysis. You can keep refining rows, columns, and prompts until the structure is ready.

Did this answer your question?