Resource generators in Recommended Actions

  • Release version: Zurich
  • Updated July 31, 2025
  • 6 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 Resource generators in Recommended Actions

    Resource generators in the Recommended Actions application provide dynamic inputs that enhance actions like guidances and field recommendations within ServiceNow workspaces such as CSM Configurable Workspace. They function as data sources that output resources—such as knowledge articles, cases, or field values—that agents can utilize to improve case resolution and decision-making processes.

    Show full answer Show less

    Resource Generator Types

    The application offers multiple types of resource generators, categorized between standard and advanced capabilities. These generators leverage logic, scripting, flows, or machine learning models to provide relevant recommendations based on the context of records.

    • Decision Table: Uses predefined input-output mappings to resolve complex decisions and generate actionable suggestions.
    • Flow: Executes subflows to automate the retrieval of recommendations like relevant knowledge base articles.
    • Scripting: Runs custom script includes to produce tailored recommendation outputs.
    • AI Search: Performs real-time AI-powered searches to find relevant records or knowledge articles based on context fields.
    • Classification (Predictive Intelligence): Applies machine learning to recommend field values during record creation by predicting probable categories or assignments.
    • Similarity (Predictive Intelligence): Finds existing records similar to the current context to provide relevant recommendations like resolved cases.
    • Similarity with Trend: Extends similarity analysis by identifying trending patterns among related records to enhance recommendation relevance.
    • Task Intelligence Classification and Similarity: Uses Task Intelligence machine learning models for flexible, automated field value predictions or recommendations, including trend-based similarity evaluations.

    Practical Use and Configuration

    Each resource generator type requires specific inputs such as solution definitions, model references, or fields to search. Customers can select from out-of-the-box generators or create custom ones to fit their unique business needs. The advanced Task Intelligence types require the Task Intelligence Admin Console plugin to be installed.

    Performance and Timeout Handling

    To ensure responsive user experience, machine learning-based resource generators operate with a 1-second timeout controlled by a system property. If prediction calls exceed this limit, they are terminated to prioritize faster system responses, and a timeout message is generated. This mechanism applies to similarity and trend-based ML generators.

    Benefits for ServiceNow Customers

    • Enhances agent productivity by automatically recommending relevant knowledge, cases, and field values.
    • Leverages machine learning to improve accuracy and relevance of recommendations.
    • Supports automation and complex decision-making through flows, scripting, and decision tables.
    • Offers customization flexibility to tailor resource generators to specific organizational requirements.
    • Maintains performance balance with built-in timeout controls for ML predictions.

    Resource generators in Recommended Actions provide resources that you can use as inputs to actions such as guidances and field recommendations.

    Resource generators provide resources such as a knowledge article, a set of cases, or a field value that you can use in recommendations. Agents can use these recommendations in the Recommended Actions panel in a workspace such as CSM Configurable Workspace.

    Another way to think of a resource generator is as a function with inputs and outputs. For example, you can create a resource generator that predicts a knowledge article and then use that knowledge article as a recommended action for a case.

    You can select a resource generator from the list or create a resource generator. For a complete list of resource generators that are included with the Recommended Actions application, see Components installed with Recommended Actions.

    Resource generator types

    The Recommended Actions application provides the following resource generator types:
    • Decision table
    • Flow
    • Scripting
    • AI search
    The Recommended Actions - Advanced application provides the following additional resource generator types:
    • Classification
    • Similarity
    • Similarity with Trend
    • Task Intelligence Classification
    • Task Intelligence Similarity
    • Task Intelligence Similarity with Trend

    These resource generator types are based on the Predictive Intelligence framework that you can use to create machine learning solutions.

    Resource generator type Definition
    Decision table Returns the results of the selected decision table. Use decision tables to help you resolve a complex decision that depends on multiple inputs and predetermined outputs. For example, a decision table might require inputs such as age, location, and health history to provide suggestions for different aspects of health insurance. You can use these suggestions, which are the outputs of the decision table, as action inputs for a recommendation.

    For more information, see Decision Tables.
    Flow Returns recommendations generated by executing the selected subflow. Use subflows to define logic and automate outputs for the resource generators. For example, you can find relevant KB articles using inputs such as product name, activation state, and publishing status. You can then propose to attach these KB articles as a recommended action.

    For more information, see Flow Designer.
    Scripting Returns recommendations generated by executing the selected script include. You must create an implementation of the ScriptingGeneratorFactory template to use a script include in the Scripting resource generator type.

    For more information, see Script includes.
    AI search Returns the AI search results and genius results based on real-time data and insights regardless of the search source. These results are relevant to the context of a record. This resource generator type requires the following inputs:
    • Search field - The field on which the AI search is performed.
    • Top N results - The number of records to return from the AI search.

    For example, you can configure AI search to recommend a KB article or similar case based on the relevancy and context of the short description. For more information, see AI Search.
    Classification Uses machine learning algorithms to recommend field values during record creation. This resource generator type requires the following inputs:
    • Classification definition - The Predictive Intelligence classification model to return the predicted values or record references for a field on a table.
    • Top N Results - The number of values or records to return from the relevant records predicted by the classification definition.

    For example, you can configure the classification solution to recommend a value for the category or assignment group based on the short description. For more information, see Predictive Intelligence.
    Similarity Identifies existing records with similar values to a context record. You must create a word corpus that acts as a vocabulary for the machine learning system to compare trained records. This resource generator type requires the following inputs:
    • Similarity definition - The similarity solution definition to return records that are relevant to the context record.
    • Top N Results - The number of records to return from the relevant records predicted by the similarity definition.

    For example, you can find similar resolved cases to propose a resolution as a recommended action. For more information, see Predictive Intelligence.
    Similarity with Trend Uses a trend definition to find commonality in the records returned by the similarity solution. You can configure the trend definition using a common condition, common field, or common reference, and a threshold value for the minimum number of records that must follow a trend.
    This resource generator type requires the following inputs:
    • Similarity definition - The similarity solution definition to return records that are relevant to the context record.
    • Trend definition - The trend definition to filter the relevant records that follow a trend.

    For example, you can find similar major incidents with a trend. Then, you can propose to link the current context record to this major incident.
    Task Intelligence Classification Uses machine learning models created in Task Intelligence Admin Console to recommend field values. The models provide flexible options to either auto-fill values on the record form or to provide recommendations depending on the sensitivity of those fields.
    Note:
    The field values are not recommended for fields when the TI model is in the background mode.

    This resource generator type requires the following inputs:

    • Solution: The solution contains configuration for recommended input fields and recommended output fields for prediction. The solution also contains preference whether the fields are auto-filled or a recommendation message is displayed. The Task Intelligence classification model predicts multiple values or record references for fields on a table.
    • Top N Results - The number of values or records to return from the relevant records predicted by the solution definition.

    For example, you can configure the Task Intelligence classification solution to recommend a value for the category or assignment group based on the short description. For more information, see Task Intelligence.

    Note:
    The Task Intelligence Admin Console (com.sn_ti_admin) plugin must be installed for selecting the Task Intelligence Classification resource generator type.

    Task Intelligence Similarity

    Identifies existing records with similar values to a context record. Uses machine learning models created in the Admin Console to recommend field values.

    This resource generator type evaluates the results of the Task Intelligence Admin Solution and returns the predicted values/records for multiple fields.

    This resource generator type requires the following inputs:
    • Model: Reference to the Task Intelligence Admin Similarity model.
    • Top N Results: The number of values or records to return from the relevant records predicted by the solution definition.
    This resource generator type returns the following outputs:
    • Predicted value: For non-reference type fields, the Task Intelligence Admin Solution returns a predicted value.
    • Predicted record: For reference type fields, the Task Intelligence Admin Solution returns a predicted record.
    For more information, see Task Intelligence.

    Task Intelligence Similarity with Trend

    This resource generator evaluates the trend on the results of TI Similarity prediction.

    Using the solution (sn_ti_admin_solution) and trend (sn_nb_action_trend_definition) definitions, this resource generator type identifies records similar to the context record. Further, it analyzes these similar records to determine if they follow the specified trend.

    This resource generator type requires the following inputs:
    • Solution: Reference to the Task Intelligence Admin Similarity Model.
    • Top N Results: Specifies the number of predicted values/records to be sent to the similarity prediction server.
    • Trend definition: Reference to the Trend Definition (sn_nb_action_trend_definition) table.
    This resource generator type returns the following outputs:
    • Count: Number of total records found in the trend.
    • Oldest created record: Oldest record in the trend.
    • Highest confidence record: Reference of the highest confidence ML result record.
    • Qualified Records Sys IDs: List of sys IDs of all qualified records.
    For more information, see Task Intelligence.

    Timeout handling for resource generators

    The Recommended Actions application uses a subflow API with a defined timeout when calling Machine Learning (ML) resource generators. This timeout ensures the RA generation engine prioritizes faster response times by terminating stalled ML prediction calls.

    The subflow has a timeout setting of 1 second. If there is a delay and the predict call times out, the system generates a message.

    The subflow applies to the following ML-based resource generators:
    • Similarity
    • Task Intelligence Similarity
    • Similarity with Trend
    • Task Intelligence Similarity with Trend

    The sn_nb_action_adv.similarity_prediction.timeout system property controls the timeout value of the subflow. The default value is set to 1000ms.