April 5, 2022 5 min read

Power Automate Desktop: Robotic Process Automation for Everyone

Power Platform Power Automate RPA Desktop Automation Low-Code

Power Automate Desktop brings robotic process automation (RPA) to Windows, enabling automation of legacy applications, web interfaces, and desktop processes without code.

Getting Started

Power Automate Desktop is free for Windows 10/11 users. Premium features require a Power Automate license.

Core Concepts

components:
  flows: Automation sequences
  actions: Individual steps (400+ built-in)
  variables: Data storage during execution
  subflows: Reusable flow components
  error_handling: Try-catch blocks
  loops: Iteration constructs

Building Desktop Flows

UI Automation

# Record or build UI interactions

# Launch application
System.RunApplication ApplicationPath: 'C:\Program Files\App\app.exe'

# Wait for window
UIAutomation.WaitForWindowToBeReady Window: MainWindow Timeout: 30

# Click button
UIAutomation.Click Element: SubmitButton ClickType: LeftClick

# Enter text
UIAutomation.PopulateTextField TextField: UsernameField Text: '%Username%'

# Read text
UIAutomation.GetElementText Element: ResultLabel TextValue=> ExtractedText

Web Automation

# Browser automation

# Launch browser
WebAutomation.LaunchChrome Url: 'https://app.example.com' BrowserInstance=> Browser

# Wait for element
WebAutomation.WaitForElementOnWebPage Browser: Browser Selector: 'input[name="search"]' Timeout: 30

# Enter text in web field
WebAutomation.PopulateTextField Browser: Browser TextField: 'input[name="search"]' Text: '%SearchTerm%'

# Click web element
WebAutomation.Click Browser: Browser Element: 'button[type="submit"]'

# Extract web data
WebAutomation.ExtractDataFromWebPage Browser: Browser Selector: 'table.results' ExtractedData=> DataTable

# Close browser
WebAutomation.CloseWebBrowser Browser: Browser

Excel Automation

# Excel operations

# Launch Excel
Excel.LaunchExcel Visible: True Instance=> ExcelInstance

# Open workbook
Excel.OpenWorkbook Instance: ExcelInstance Path: 'C:\Data\Report.xlsx' Workbook=> Workbook

# Read range
Excel.ReadFromExcel Instance: ExcelInstance Range: 'A1:D100' ReadAsText: False DataTable=> ExcelData

# Write data
Excel.WriteToExcel Instance: ExcelInstance DataToWrite: %ProcessedData% StartColumn: 1 StartRow: 1

# Run macro
Excel.RunMacro Instance: ExcelInstance Macro: 'FormatReport'

# Save and close
Excel.SaveExcel Instance: ExcelInstance
Excel.CloseExcel Instance: ExcelInstance

Variables and Data

Working with Variables

# Set variable
Variables.SetVariable Name: 'Counter' Value: 0

# Increment
Variables.IncreaseVariable Name: 'Counter' Value: 1

# Create list
Variables.CreateNewList List=> ItemList

# Add to list
Variables.AddItemToList List: ItemList Item: '%CurrentItem%'

# Create data table
Variables.CreateNewDataTable Columns: ['Name', 'Email', 'Status'] DataTable=> ResultTable

# Add row to data table
Variables.AddRowToDataTable DataTable: ResultTable RowData: ['John', 'john@example.com', 'Active']

Conditions and Loops

# If condition
IF %Status% = 'Pending' THEN
    # Actions for pending status
    UIAutomation.Click Element: ApproveButton
ELSE IF %Status% = 'Review' THEN
    # Actions for review status
    UIAutomation.Click Element: ReviewButton
ELSE
    # Default actions
    Text.AppendLine Text: %Log% LineToAppend: 'Unknown status: %Status%'
END

# For each loop
LOOP FOREACH CurrentRow IN %DataTable%
    # Process each row
    Variables.SetVariable Name: 'CustomerName' Value: %CurrentRow['Name']%

    # Perform actions
    UIAutomation.PopulateTextField TextField: NameField Text: '%CustomerName%'

    # Handle errors
    ON BLOCK ERROR
        Text.AppendLine Text: %ErrorLog% LineToAppend: 'Error processing: %CustomerName%'
    END
END

# While loop
LOOP WHILE %Counter% < 100
    # Increment counter
    Variables.IncreaseVariable Name: 'Counter' Value: 1

    # Check condition to break
    IF %FoundItem% = True THEN
        EXIT LOOP
    END
END

Error Handling

# Try-catch pattern
BLOCK 'Process Invoice'
ON BLOCK ERROR
    # Log error
    Text.AppendLine Text: %ErrorLog% LineToAppend: 'Error: %LastError%'

    # Take screenshot for debugging
    System.TakeScreenshot Screen: 0 File: 'C:\Logs\error_%DateTime%.png'

    # Send notification
    Email.SendEmail Server: 'smtp.company.com' To: 'support@company.com' Subject: 'RPA Error' Body: '%LastError%'

    # Decide whether to continue or stop
    IF %ErrorCount% > 3 THEN
        THROW EXCEPTION 'Too many errors'
    END
END
    # Normal processing
    UIAutomation.Click Element: SubmitButton
    UIAutomation.WaitForElement Element: SuccessMessage Timeout: 30
END

Integration with Cloud Flows

Trigger Desktop Flow from Cloud

{
    "trigger": {
        "type": "Request",
        "inputs": {
            "schema": {
                "type": "object",
                "properties": {
                    "invoiceNumber": {"type": "string"},
                    "amount": {"type": "number"}
                }
            }
        }
    },
    "actions": {
        "Run_Desktop_Flow": {
            "type": "ApiConnection",
            "inputs": {
                "host": {
                    "connection": {
                        "name": "@parameters('$connections')['shared_uiflow']['connectionId']"
                    }
                },
                "method": "post",
                "path": "/providers/Microsoft.ProcessSimple/environments/@{encodeURIComponent('Default')}/flows/@{encodeURIComponent('desktop-flow-id')}/runs",
                "body": {
                    "invoiceNumber": "@triggerBody()?['invoiceNumber']",
                    "amount": "@triggerBody()?['amount']"
                }
            }
        }
    }
}

Return Data to Cloud Flow

# In desktop flow
# Process data and prepare output

# Set output variable (defined in flow properties)
Variables.SetVariable Name: 'OutputResult' Value: %ProcessedData%
Variables.SetVariable Name: 'OutputStatus' Value: 'Success'

# These variables are returned to the cloud flow

OCR and Document Processing

# OCR for image text extraction

# Capture region
System.TakeScreenshotOfRegion Region: %DocumentArea% Image=> CapturedImage

# OCR extraction
OCR.ExtractTextWithEngine Image: %CapturedImage% Engine: WindowsOCR Text=> ExtractedText

# Parse extracted text
Text.RegexMatch Text: %ExtractedText% Pattern: 'Invoice #: (\d+)' Match=> InvoiceNumber

# PDF text extraction
PDF.ExtractTextFromPDF File: 'C:\Documents\invoice.pdf' Text=> PDFText

Best Practices

Reliable Selectors

# Use multiple selector attributes for reliability
UIAutomation.WaitForElement
    Element: {
        "Name": "Submit",
        "ClassName": "btn-primary",
        "AutomationId": "submitButton"
    }
    Timeout: 30
    FoundElement=> SubmitBtn

# Add fallback selectors
IF %SubmitBtn% = '' THEN
    UIAutomation.WaitForElement
        Element: {"Name": "Submit Order"}
        Timeout: 10
        FoundElement=> SubmitBtn
END

Performance Optimization

tips:
  - Use keyboard shortcuts over mouse clicks
  - Minimize screen captures
  - Close applications when done
  - Use specific waits, not fixed delays
  - Batch operations when possible
  - Disable UI animations if possible

Maintenance

# Add logging throughout
Text.AppendLine Text: %Log% LineToAppend: '[%DateTime%] Starting process'

# Version control your flows
# Export flows regularly
# Document selector dependencies
# Test after application updates

Unattended vs Attended

attended:
  description: User interacts with desktop during execution
  use_cases:
    - User-triggered tasks
    - Assisted automation
    - Training scenarios

unattended:
  description: Runs on dedicated machine without user
  use_cases:
    - Scheduled batch processing
    - Background operations
    - High-volume processing
  requirements:
    - Windows machine (VM or physical)
    - Power Automate unattended license
    - Machine registered in Power Platform

Conclusion

Power Automate Desktop democratizes RPA:

No-code automation for legacy apps
Integration with cloud flows
Built-in OCR and document processing
Free for basic Windows automation

It’s particularly valuable for automating processes in applications without APIs.