Column Extraction
The Column Extraction feature enables you to automatically extract structured data fields from unstructured sources — such as meeting transcripts, recorded calls, or text documents — using an AI agent. Rather than manually reviewing lengthy content, you can configure an extraction pipeline that pulls specific fields into a structured workbook with minimal effort.
Overview
When unstructured records (e.g., meeting transcripts) are imported into a workbook, the extraction feature allows you to define the fields you want to capture, preview the results, save the configuration for reuse, and run the extraction across multiple records simultaneously.
Watch a Quick Demo
Prerequisites
Before using Column Extraction, ensure the following:
- Your source data (e.g., transcripts) has been imported into a workbook either through import or query.
- You have at least one record available to extract from.
- You have access to an existing destination workbook, or you are prepared to create a new one.
Configuring an Extraction
Step 1: Accessing the Extraction Tool
- Open your workbook containing the source records.
- Select the record(s) you wish to extract from.
- Click the Extract option from the dropdown menu on the column toolbar.
The extraction panel will open, displaying the selected source record on the left and configuration options on the right.
Step 2: Choose a Destination Workbook
In the extraction panel, select where the extracted data should be written:
| Option | Description |
|---|---|
| Create a New Workbook | Generates a new workbook with columns based on the fields you define. |
| Use an Existing Workbook | Maps extracted fields to columns that already exist in a chosen workbook. |
Step 3: Name the Extraction
Enter a descriptive name for this extraction configuration (e.g., Financial Extraction).
Note: Naming the extraction is important. Saved extractions can be reused when new records are added, eliminating the need to reconfigure from scratch each time.
Step 4: Select an Extraction Mode
Choose the mode that best fits the structure of the data you need to extract:
| Mode | Best For |
|---|---|
| Standard | Simple, flat fields — such as names, numbers, or single values. |
| Advanced | Nested objects, complex structures, or data requiring a specific output format. |
Step 5: Define the Fields to Extract
In the prompt input, describe the fields you want the AI agent to extract. For example:
Extract: NRR, gross margin percentage, burn rate, fundraising amount, and lifetime value.
Click Generate. The agent will process the prompt and prepare the corresponding extraction fields automatically.
Adding Fields Manually
After generation, you can add additional fields at any time:
- Click Add a Field.
- Provide a column name for the new field.
- Enter a description to instruct the agent on what to extract.
- If using an existing workbook, map the field to an existing column.
Step 6: Set a Primary Key
Select the column to use as the primary key for the destination workbook (e.g., an ID column). This uniquely identifies each extracted record.
Step 7: Previewing Extraction Results
Before saving and running an extraction, you can validate the output:
- Click Preview Results.
- The agent will run the extraction on the currently selected record only.
- Review the extracted values in the preview panel to confirm all fields are populating correctly.
Tip: Always preview your extraction on at least one representative record before saving the configuration, especially when working with advanced mode or nested fields.
Step 8: Saving the Extraction Configuration
Once the fields and settings are verified:
- Confirm or update the configuration name.
- Choose whether to store extracted data as JSON in addition to structured columns.
- Enabling this option preserves the raw AI response alongside the structured output.
- This is useful if you need to audit or reprocess responses at a later time.
- Click Create to save the extraction configuration.
Step 9: Running an Extraction
After saving, the extraction is ready to execute:
- Click Run.
- Select the records you wish to process.
- Review the processing option:
- Process New Cells Only (enabled by default) — Skips records that have already been extracted, preventing duplicate processing.
- Click Run Extraction to begin.
A progress screen will display the status of the extraction job. Once complete, a success notification will appear in the top-right corner of the interface.
Step 10: Viewing Extracted Data
Upon completion:
- Click View in the success notification.
- The destination workbook will open, displaying all extracted fields as structured columns across your records.
Reusing a Saved Extraction
Saved extraction configurations are stored and accessible for future use. When new records or transcripts are added to your source workbook:
- Navigate to the saved extraction configuration.
- Click Run.
- Select the new records (or use Process New Cells Only to auto-filter).
- Run the extraction — no reconfiguration required.
Was this helpful?