Limitations of Robotic Process Automation (RPA) bot generation skill

  • Release version: Xanadu
  • Updated January 30, 2025
  • 3 minutes to read
  • Summarize
    Summarized using AI
    This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.

    Summary of Limitations of Robotic Process Automation (RPA) bot generation skill

    The RPA bot generation skill in ServiceNow leverages AI Search and large language models (LLM) to interpret user instructions and create automation workflows. However, there are several limitations affecting its current functionality and usability, particularly in the Xanadu and Yokohama release versions.

    Show full answer Show less

    Key Limitations

    • API Timeouts and Rate Limits: The RPA bot generation skill is subject to the timeout and rate-limiting constraints of AI Search and LLM APIs, which can impact response times and workflow generation.
    • Component Scope Restriction: The initial version supports a limited set of components. Components outside this predefined scope (e.g., Capture component for screenshots) do not appear in the generated workflows, causing gaps in automation tasks.
    • Plugin Version Compatibility: The feature relies on Yokohama version plugins. Using plugins from other versions (e.g., Washington DC) may result in methods not being recognized, leading to "unknown components" in workflows. This incompatibility can affect connectors like Universal App and Data Reader components.
    • Use of Existing Connectors: When creating new automation activities, the system reuses the most recent global object of the same connector type to maintain consistency, which is important to understand when multiple instances of connectors like Microsoft Excel are in use.
    • Lack of Dynamic Configuration: Yokohama release does not support dynamic configuration for components, requiring manual input for configuration parameters such as those in the Microsoft Excel connector's SetCellValues method.
    • Automation for Windows, Java, and Browser Applications: For these applications, the system suggests generic UI actions (SetValue, GetValue, Click) from the Universal app connector. Users must manually configure these actions, including capturing screens and adding elements.
    • Expression-Based Components: Components that require expressions (e.g., Decision or Expression components) do not have expressions automatically populated, necessitating manual input.
    • Hallucinated Components: If the LLM returns methods that do not exist in RPA Desktop Design Studio, these appear as unknown components in the generated workflow, which may require manual correction or adjustment.

    Practical Considerations for ServiceNow Customers

    Understanding these limitations helps set realistic expectations when using the RPA bot generation skill. Customers should anticipate manual configuration for certain components and verify plugin compatibility to avoid unknown components in their workflows. Awareness of API limits and supported components ensures smoother automation workflow creation and reduces troubleshooting time. This knowledge is essential for planning and scaling RPA initiatives effectively within the ServiceNow platform.

    The limitations of the RPA bot generation skill are as follows.

    AI Search and LLM timeout
    The RPA bot generation skill relies on AI Search and large language models (LLM) to understand the user instructions and flow creation. The timeout limits and rate-limiting behavior of these APIs collectively apply to the RPA bot generation skill. For more information about AI search system properties, see AI Search system properties. For more information about API rate limits, see Inbound REST API rate limiting. For more information about configuring rate limiting for providers, see Configure rate limiting for providers.
    Component scope in Yokohama release
    The initial version supports few components. Components outside this scope don't appear in the generated automation workflow. For example, if you provide an instruction, such as Capture a screen shot and save it in the location at 'C:\Users\Screens', the generated automation workflow might not contain the Capture component because it isn’t included in the initial version components list. For more information about the list of supported components, see Supported components and connectors for Now Assist for RPA Hub in Yokohama release.
    Plugin version compatibility
    The feature uses Yokohama version plugins. If the large language model (LLM) returns a method that doesn’t exist in the loaded plugin version, then such methods are displayed as unknown components in the generated automation workflow. For example, if you’re using Washington DC version plugins within an existing automation project and provide an instruction to set or get values from an application, then the SetValue, GetValue, or Click methods of the Universal App connector are displayed as unknown components in the generated automation workflow.
    In the following example, to get the weather information, data reader components are required. Since they are not available in the current scope, they appear as UNKNOWN components in the automation preview.
    Figure 1. Unknown components in the Automation preview
    Automation preview section displays UNKNOWN components.
    If you select Next in the previous screen, the automation appears as follows on the Design surface.
    Figure 2. Unknown components on the Design surface
    Design surface displays UNKNOWN components.
    Usage of existing connectors in new activity using Now Assist or build automation context menu options
    During the automation workflow creation, if a method, such as the AddNewWorksheet method of Microsoft Excel connector is created, RPA Desktop Design Studio checks for an existing global object of the same connector type and uses the most recent one. In the following example, the Badging Data Microsoft Excel connector is used.
    Figure 3. Microsoft Excel connectors in Global Objects
    Project Explorer section displays Microsoft Excel connectors in the Global Objects. Badging Data Microsoft Excel connector is selected.
    Dynamic configuration
    Dynamic configuration for supported components isn’t available in the Yokohama release. You can configure components manually. For example, you must give inputs in the configuration window for the Microsoft Excel connector's SetCellValues method. Select the settings icon (Settings icon.) to view the configuration window.
    Data in the parameter values might not be populated in the large language model (LLM) generated components.
    Figure 4. Configuration window for SetCellValues method
    Configuration window for the SetCellValues method is displayed.
    Automation for Windows, Java, or browser applications
    For automating Windows, Java, or browser-based applications, RPA Desktop Design Studio suggests generic UI actions such as SetValue, GetValue, or Click. These actions are a part of the Universal app connector. You must configure these actions manually, including capturing screens and adding elements via the field window. For more information about these methods, see Universal app connector.
    Expression-based components
    For components that require expressions, such as Decision or Expression, and so on, expressions in these components aren’t automatically populated in the Xanadu release.
    Hallucinated components
    If the large language model (LLM) returns a method that doesn’t exist in the RPA Desktop Design Studio, then such methods are displayed as unknown components in the generated automation workflow.