In this blog, you will learn how Agentic Desktop brings intelligent, no-code automation to your local environment. It enables you to automate repetitive, UI-based tasks in legacy desktop applications without APIs or backend integrations.
How is it useful?
The Agentic Desktop application is used to create desktop agents in the ServiceNow AI Platform to interact with browsers, thick-client applications, and legacy systems and automate end-to-end workflows using AI-leveraged desktop actions.
This is useful because it reduces time spent on repetitive, manual desktop tasks by enabling no-code automation in applications that lack APIs or backend integrations. By allowing users not just developers to design and run desktop actions through intuitive workspaces, Agentic Desktop helps teams complete desktop-based processes faster and more consistently.
How are desktop actions used as tools by AI agents?
Desktop actions are added to AI agents as tools in AI Agent Studio. These actions are reusable blocks that simulate user interface interactions or perform non-UI system operations such as tasks in Excel, Outlook, file management, PowerShell, or SSH. During execution, AI agents select and invoke the appropriate desktop action as a tool to carry out steps on the user's desktop as part of an end-to-end workflow.
How to Create a Desktop Action in AIA Studio: A Step-by-Step Guide
Desktop Actions are the building blocks of agentic automation on the ServiceNow platform. Whether you are automating a repetitive enterprise workflow or enabling AI agents to interact with any desktop application, the journey starts here in AIA Studio.
This guide walks you through creating your first Desktop Action from scratch, step by step.
Before You Begin
Make sure you have all the pre-requisites in place if not then click on the link to ensure you have access to Agentic Desktop app.To know more on the set-up refer :Agentic Desktop pre-requisites
- Access to AIA Studio on your ServiceNow instance (Vancouver or later)
- Agentic Desktop App installed and role assigned to your user (sn_aia.admin)
- A target desktop application installed and accessible on your execution environment that you intend to automate
- Basic familiarity with ServiceNow flows or process automation concepts
Step 1: Open Agentic Desktop
- Launch the application installed on your Windows machine.
- Enter your ServiceNow instance URL (e.g., https://instance.service-now.com).
- Select Proceed UI action
Step 2: Sign in to ServiceNow Instance
- Enter the credentials of the user account that has the sn_aia.admin role to log in.
Step 3: Create desktop action
- Select "Create Desktop Action" on the landing page. After selecting "Create Desktop Action," the New Desktop Action dialog box appears. By default, the Auto-capture steps option is enabled.
Step 4: Details of the desktop action
Give your Desktop Action a clear identity so your team and your AI agents know exactly what it does.
- Action Name: Use a descriptive name that reflects the task e.g., "Submit Expense Report in SAP".
- Description: Write one or two sentences describing what the action does and when it should be used. This surfaces in agent selection and search.
Step 5: Capturing method
- Select Auto capture steps option (recorder) and select start capturing
Here we will leverage the capability of Recorder. With the recorder feature, you can record your interactions with desktop applications to create automated workflows. It records every step you take, including clicks, keystrokes, and data entry. Let us quickly explore how to record steps using this method.
Step 6: Start recording
- The Agentic Desktop window minimizes, and the recorder panel appears on the top right of the screen. Open the desktop application where you want to capture steps. For our demonstration, we have selected the Employee Badge Management application. Then, select "Start" on the recorder panel to begin capturing.
Step 7: Record and Stop
- Perform the steps you want to capture. You can also pause or discard the recording if necessary. When finished, select "Stop" on the recorder panel to complete the capture. Guiding Principles of Recording in Agentic Desktop
Step 8 : Design Workspace
- The recorded steps are visible as screenshots in the Design workspace, with anchors and steps automatically assigned. You can fine-tune the Anchors and Steps if required.
- Naming of Screens, Anchors, and Steps - provide names for all screens, anchors, and steps on the properties panel, right side of the screen. You can edit the default names that are auto-generated. Remember that the names should only have alphanumeric characters (no spaces or special characters). Also, the names must be unique at their respective parent level.
For more information about screen, anchor, and step properties, refer to the Product Documentation website.
Step 9 : Details Tab
- Once you are done capturing the screens and steps, select the Details tab. Here you can edit the name and description fields if required.
- Select the applications you want to automate from the Application list. The list displays the applications available on the ServiceNow instance. If an application is not listed, you can add it by selecting the Add (+) icon.
- In the Input and Output Parameters section, review the parameters related to steps for each screen. Finally, select "Save" to complete the process.
Step 10: Test Your Desktop Action
- Never skip testing. A Desktop Action that runs well in a demo environment may behave differently under real conditions.
- Click the Test UI action, a Test modal opens.
- Enter sample values for each input parameter
- Click Run Test and observe the execution in the launched Execution Workspace.
- Review the execution
- Iterate on your steps or anchor configuration until the action completes successfully end to end.
Step 11: Activate and Make Available to Agents
Once your action is tested and stable, it is time to Activate/publish it so AI agents can discover and execute it.
- Click Activate from the Desktop Action record.
- Assign the action to the relevant AI Agent configuration.
- Set execution permissions specify which agents or flows can invoke this action.
Step 12: Add desktop action to AI Agent
To enable desktop automation, you need to add a desktop action as a tool to your AI agent in AI Agent Studio.
- Navigate to the Add tools and information section. Select Desktop Action from the Add tool drop-down list.
Refer to the product documentation to learn more about how to create an AI agent.
Step 13: Add desktop action
- The Add a desktop action dialog box appears. Complete the form with the relevant information.
Step 14: UI Block selection
UI block actions simulate human UI steps (click, type, select, etc.). Use these when the target application has no available API. If this option is selected, then choose a desktop action from the list (only activated actions appear in it).
- Select the relevant desktop action that you have created then several read-only fields appear (Desktop action description, Applications, Created by, Last updated, Inputs, and Outputs).
Step 15: Provide name and description
- Provide a Name for the new desktop action and a detailed description of the tool. Explain its purpose and specific functionality, including how the AI agent uses it to perform tasks.
For a detailed explanation of each field in this modal, refer to this link.
- Select Add to complete the process.
Step 16: Trigger from Now Assist Panel
- Navigate to the Home page and open the Now Assist panel. However, ensure that you have enabled the Display toggle in the Now Assist panel settings to allow the agentic workflow to surface in the panel. And you have selected Save, then select Test to complete the agentic workflow creation. To know more about how to create workflow in AIA studio refer to the link
- Enter a prompt that clearly provides all the inputs required for the automation to complete.
- Before triggering execution, ensure popups are enabled in your browser, as the Execution Workspace launches in a new window.
- Submit the prompt. The automation starts, triggering the badge management workflow, which in turn triggers the AI agent.
- The agent prepares a plan and displays which Desktop Action it will use for execution.
- Select Open when the Pop- up of "This site is trying to open Agentic Desktop" appears on top of the window. The Execution Workspace launches.
- Enter your Windows Security credentials to log in. You will only be prompted for credentials once per day.Once logged in, minimize the window rather than closing it to avoid restarting the background session for future executions.
- Monitor the automation as it runs in the Execution Workspace.
- Once complete, the Now Assist panel displays the outcome of the automation as Success
What Is Next?
Now that your first Desktop Action is live, here are a few ways to go deeper:
- Build a library of reusable actions organized by application and business process
- Got a use case in mind? Share it in the comments or on the ServiceNow Community. The best submissions could be featured as an out-of-the-box Desktop action use case .🙂
-
To build a strong foundation before diving deeper, we recommend completing the Agentic Desktop Fundamentals course available on ServiceNow University. It covers the core concepts, architecture, and hands-on essentials you need to get the most out of the Agentic Desktop platform.
