Extract information using AI

This is a preview release of Gemini in AppSheet Solutions. See Track and manage feature releases. Preview offerings are intended for use in test environments only. This feature is not recommended for use in production apps. For questions or assistance with this feature, contact AppSheet Support.

Gemini in AppSheet Solutions (preview) requires an AppSheet Enterprise Plus account.
 

Extract information from an uploaded image (such as a photo) or a file (PDF) using AI by adding the Extract AI task in an automation.

See also:

Note: For the preview release, the Extract AI task can extract information from images and PDF files only.

Get started

Step through a quick start to build your first automation using AI to extract information from a car photo and categorize the car by body style. See Quick start: Build your first automation using AI.

Add and configure the Extract AI task

To extract data from an uploaded image (such as a photo) or a file (PDF), add and configure the Extract AI task by performing the following steps.

For the preview release, you must turn on Gemini in AppSheet Solutions in each app where you want to use AI in automations.
  1. Add the Extract AI task to a bot, as described in Add a task to a bot.
  2. Configure the Extract AI task by using the settings described in the following table. 
  3. Save the app.

Setting

Description

Task type

Select AI task.

Table

Table defined in the event and used by the task. See Tables: The Essentials for more information about tables.

AI task 

Type of AI task. Select Extract.

Input column

Column in the table that is used as input to the AI task.

For the Extract task, this must be one of the following types: Drawing, File, Image, or  Thumbnail 

Note: For the preview release, the Extract AI task can extract information from images and PDF files only.

To view or edit the structure of the selected column, position your cursor in the field and click Edit icon  to configure the column properties.

Output

Select one of the following:

  • Save to table to save the extracted information to the table in your data source.
  • Return value to guide AI responses without storing the results in your data source.

Then, select one or more columns to define the information to extract, up to a maximum of 20. Select a column from the drop-down. Click Add to add more columns.

To save the extracted information to the table in your data source, select Save to table. Otherwise, select Return value to capture the values returned with storing them.

Notes: 

For best practices when saving content, see Manage updates to row data by multiple bots.

Additional instructions

Recommended. Do the following:

  • Describe in more detail the information you want to extract.
  • Indicate what values to set if results are unknown. 
  • Ensure that any recommended values match the type and format of the column, for best results. 

Supports up to 1000 characters. 

For example:  

Extract the make, model, color, and year
If not sure of Make or Model, set to "Unknown"
If not sure of Color, set to "Other"
If not sure of Year, set to "0"

 

Was this helpful?

How can we improve it?

Need more help?

Try these next steps:

Contact us Tell us more and we’ll help you get there
true
Search
Clear search
Close search
Main menu
3181032003260194764
true
Search Help Center
true
true
true
false
false
false
false