Quick start: Extract information from a PDF file using AI

Using AI in automation, the process can be greatly simplified for the end user. Here’s what the experience could be for them:

1. Receive a service ticket request as a PDF document.
2. Upload the PDF document to your AppSheet app.

Using AI, AppSheet extracts data from the PDF file, such as title, ticket ID, description, and incident date.

In this Quick start, you'll build an automation that uses AI to open a service ticket request. Specifically, you'll create an AppSheet database and use it to create an app, then build an automation using AI to extract information from a PDF file and open a service ticket that specifies a severity rating.

To build an automation to extract information from a PDF file and open a service ticket request, do the following steps:

1. Create an AppSheet database
2. Create an app using the AppSheet database
3. Build an automation to extract information from a PDF file using AI
4. Run the automation in the app preview

Step 1: Create an AppSheet database

First, you'll  create an AppSheet database and add the columns you need to store information about the service ticket request, such as title, ticket ID, description, incident date, and so on.

To create an AppSheet database, do the following:

1. Sign in to AppSheet.
   The My Apps page is displayed.
2. Select Create > Database > New database.
   A new AppSheet database is created and opened in the database editor
3. Click within the name, Untitled database, and change the name of the database to: Service request
   
   
4. Change the name of the Table 1 table:
   1. Position your cursor in the Table 1 tab and select More  > Table settings.
      
      
   2. Change the table name to: Service request
   3. Leave all other fields set to the default values.
   4. Click Save.
5. Delete the dummy rows in the table, as follows:
   1. Click in column 1 and drag the cursor to highlight all rows in the table. 
      
      
   2. Right-click on the selected rows and select Delete rows.
      
      
      A confirmation dialog displays to confirm the delete operation.
   3. Click OK to confirm the operation.
6. Edit the column Assignee column by doing the following:
   1.  Position your cursor in the Assignee column header and select More  > Edit column.
   2. Change the Name field to Ticket ID.
   3. Change the Type field to Text.
   4. Click Save.
      
      
7. Add a new Description column, as follows:
   1. Click + Add column in the column header.
      
      
   2. Change the Name field to Description.
   3. Set the Type field to LongText.
   4. Click Save.
      
      
8. Add a new Severity column, as follows:
   1. Click + Add column in the column header.
   2. Change the Name field to Severity.
   3. Set the Type field to Enum.
   4. Set the Item type field to Dropdown.
   5. Click in the first option field and enter Low.
   6. Click Add option, enter Medium, and set the color to white. 
   7. Click Add option, enter High, and set the color to white. 
   8. Click Save.
      
      
9. Add a new Document column, as follows:
   1. Click + Add column in the column header.
   2. Change the Name field to Document.
   3. Set the Type field to Attachments > File.
   4. Click Save.
      
      

The columns in the AppSheet database appear as shown:

Next, you'll create an app using the AppSheet database.

 Step 2: Create an app using the AppSheet database

To create an app using the AppSheet database, do the following:

1. In the top-right corner of the database editor, click Apps.
   
   
   
   The Apps using Service request pane displays. No apps are listed.
2. Click New AppSheet app.
   The AppSheet app is created and opened in the app editor in a new tab.
3. In the app editor, in the left navigation, click  to display the Data pane (if it isn't already displayed).
   
   
   
   The Service request table is displayed. Notice that the source is the Service request table in the AppSheet database that you created in the previous step.
   
   

Next, you'll build an automation to extract information from a photo using AI.

 Step 3: Build an automation to extract information from a PDF file using AI

To build an automation to extract information from a photo of a car using the Extract AI task , do the following:

1.  In the app editor, go to Automation .
   
   
2. Click + in the top header of the Bots pane.
   
   
3. In the Add a new bot dialog, click Create a new bot.
   
   
   
   A new empty bot is created and displayed in the center pane of the app editor.
4. Edit the name of the bot.
   1. Position your cursor over the New Bot in the left navigation pane.
   2. Select More  > Rename.
      
      
   3. Rename the bot to Track service request and press Enter.
5. In the new bot flow in the center pane, click Configure event.
   
   
6. For Event name, enter Track service request and click Create a new event.

   

   The Settings pane opens in the right side of the app editor.

7. Configure the event by defining a condition to ensure that a document file is uploaded before triggering the automation. as follows:
   1. Click in the Condition field.
      The Expression Assistant opens.
   2. In the text box enter: ISNOTBLANK([Document])
      
      
      
      This expression ensures that Document column is not blank and contains a PDF file.
   3. Click Save.
   4. Leave all other settings set to the defaults.
8. In the bot flow in the center pane, click + Add a step.
   
   
9. For Step name, enter Extract from PDF and click Create a new step.
   
   
   
   The Extract from PDF task is added and the configuration settings are displayed in the right pane.
   Note: You might need to select the task in the center pane to view the configuration settings.
10. Edit the task configuration settings in the right pane, as follows: 
    1. Set the task type set to AI task.
    2. Leave the AI task field set to Extract. This AI task type extracts information from a photo, file, or text.
    3. Leave the Input column field set to Document. This is the File column used to upload a PDF file.
    4. Under Output:
       1. Select Save to table to save the extracted information to the AppSheet database.
       2. Notice that the first column is added and set to Title. 
          Leave this column selection as is.
       3. Click Add and select Ticket ID in the drop-down. (It should be selected by default.)
       4. Click Add and select Description in the drop-down.
       5. Click Add and select Severity in the drop-down.
       6. Click Add and select Status in the drop-down.
       7. Click Add and select Date in the drop-down.
    5. In the Additional instructions field, enter:
        
       Set the title to a short sentence from the description. Set the severity to "High" for serious incidents.
       For best practices on setting the Additional instructions field, see Best practices for using AI in automations.
       The Settings pane appears as follows:
       
       
11. Click Save to save the app.
    
    

Next, you'll test your automation using the app preview.

 Step 4: Run the automation in the app preview

You can run the automation in the app preview located in the right pane.

To run the automation in the app preview, do the following:

1. Before you proceed, download the Acme Service Ticket Request - Example 1 PDF file to your local directory.
2. At the top of the right pane, click to display the mobile app preview.
   
   
3. Turn off Edit in the app preview, if necessary. This will make it easier to interact with the app in the preview pane.
   
   
4. Click + in the app preview.
   
   
5. Scroll down and click in the Document field. 
   
   
6. Navigate to the PDF file in your local drive, downloaded in step 1, and click Open.
7. Click Save.
   
   The PDF file is uploaded. AppSheet syncs the app, as indicated by the sync notification:
   
   
   
   As AppSheet syncs the app, AI extracts information from the PDF file to populate the remaining fields. When the sync completes, the service request is added to the inventory.
8. After the sync completes, click the service request in the list to view the details. 
   
   
   
   Notice that the Title, Ticket ID, Status, Description, Date, and Severity fields have been auto-populated. The Title and Severity fields are set based on your additional instructions.
   
   

Congratulations! You have built an automation using AI to extract information from a PDF. 

What's next?

• Explore more Gemini-based quick starts to learn about AI in automations.
• Explore other quick starts to learn about other AppSheet features.