> ## Documentation Index
> Fetch the complete documentation index at: https://knowledge.goautonomous.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Savvy Tables

> Extract structured data from unstructured documents by drawing and editing table boundaries directly in Flow.

## Overview

Some documents — like scanned PDFs, hand-formatted spreadsheets, or custom order forms — don't have machine-readable table structures. **Savvy Tables** lets you define table boundaries manually by drawing directly on the document, so Go Autonomous can extract the data correctly.

The process has two steps:

1. **Table Creation Mode** — draw the table boundaries.
2. **Table Edition Mode** — refine the rows and columns.

All changes you make are tracked in [History](/knowledge-base/flow/history) and can be reverted if needed.

***

## When to use Savvy Tables

Use Savvy Tables when a document contains tabular data that Go Autonomous hasn't picked up automatically — typically because the document is a scanned image, a PDF without an embedded text layer, or a custom layout where the table structure isn't machine-readable.

If the table is already being extracted correctly, you don't need to draw it manually. Savvy Tables is for the cases where the automatic extraction misses rows, merges columns, or fails to detect the table at all.

## What counts as a table

For Savvy Tables to extract data correctly, the area you draw needs to actually behave like a table. The minimum requirements are:

* **A header row** — one row at the top that labels each column (e.g. *Product Number*, *Quantity*, *Unit Price*).
* **At least one column** — every column must sit under a header. Data without a header cannot be mapped to a field.
* **Aligned rows** — each row should contain values that line up with the columns above. Free-flowing paragraphs, mixed layouts, or merged cells that span multiple columns won't extract reliably.

**What doesn't count as a table:**

* A list of values without column headers (e.g. a bullet list of product numbers).
* Two columns of unrelated data side by side without a clear header relationship.
* Free text that visually resembles a table but lacks consistent columns and headers.

If the data you need doesn't fit these rules, Savvy Tables won't help — that content needs to be handled through other extraction methods or manual entry.

***

## Step 1: Table Creation Mode

Table Creation Mode lets you draw rectangles over the areas of the document that contain table data.

<Steps>
  <Step title="Activate Table Creation Mode">
    Click the **Table Creation Mode** button (grid icon) in the left toolbar. The cursor changes to indicate you're in drawing mode.
  </Step>

  <Step title="Draw a table boundary">
    Click and drag over the area of the document that contains the table. A rectangle appears around the selected region.
  </Step>

  <Step title="Apply to multiple pages (optional)">
    If the same table structure appears across multiple pages — for example, a repeating line-item table — you can choose to apply the drawn boundaries to additional pages at once.
  </Step>

  <Step title="Save">
    Press **Ctrl+S** or click the **Save** button to confirm the table boundaries. Go Autonomous will use the drawn boundaries to structure the data for extraction.
  </Step>
</Steps>

***

## Step 2: Table Edition Mode

Once a table is drawn, use Table Edition Mode to fine-tune the row and column structure.

<Steps>
  <Step title="Activate Table Edition Mode">
    Click the **Table Edition Mode** button (pencil icon) in the left toolbar. Existing table boundaries become editable.
  </Step>

  <Step title="Add or remove rows and columns">
    Right-click inside the table to open the context menu. From here you can:

    * **Add row** — insert a new row above or below the selected position.
    * **Delete row** — remove the selected row.
    * **Add column** — insert a new column to the left or right.
    * **Delete column** — remove the selected column.
  </Step>

  <Step title="Resize rows and columns">
    Click and drag the divider lines between rows or columns to adjust their size. This is useful when the drawn boundaries don't perfectly align with the source document's layout.
  </Step>

  <Step title="Label every column">
    Each column shows its mapped field as a label above the table. Columns that haven't been mapped yet are marked **Unlabeled** with a grey badge — easy to spot against the autonomous-coloured badges of mapped columns. Data in unlabeled columns is not extracted, so assign a field to every column you want to keep before saving.
  </Step>

  <Step title="Save">
    Press **Ctrl+S** or click the **Save** button when you're satisfied with the table structure.
  </Step>
</Steps>

***

## Tips for best results

* Draw boundaries as tightly as possible around the actual table content — avoid including titles or footnotes in the rectangle.
* If a table spans multiple pages with consistent column alignment, use the **apply to multiple pages** option to save time.
* Use Table Edition Mode to clean up any rows or columns the drawing didn't capture perfectly.
* After saving, reprocess the request from the [Flow Menu](/knowledge-base/flow/menu-bar) to apply the updated table structure to extraction.

***

## What's next

* [The Flow toolbar](/knowledge-base/flow/menu-bar) — learn about all the other tools available when reviewing a request.
* [Review and confirm orders](/knowledge-base/flow/review-and-confirm) — complete the request workflow after extracting table data.