--- sourceDocument: Yokohama Build workflows sourceDocumentLink: https://servicenow-prod.fluidtopics.net/r/yokohama/build-workflows Release : - yokohama ft:locale : - en-US ft:publication_title : - Yokohama Build workflows ft:clusterId : - crworkflow bundleId : - crworkflow workflow : - Creator --- # Call Now Assist Skill step # Call Now Assist Skill step {#ariaid-title1} Release version: Yokohama Updated April 10, 2025 ![](https://www.servicenow.com/docs/portal-asset/ico-clock) 4 minutes to read Summarize ![AI sparkle icon](https://servicenow.com/docs/portal-asset/ai-sparkle-icon) 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 Call Now Assist Skill step The Call Now Assist Skill step enables ServiceNow customers to run a published Now Assist skill directly from an action within Workflow Studio. This integration allows you to configure skill inputs and outputs dynamically using step inputs and outputs, facilitating AI-powered automation and extended capabilities within your workflows. Show full answer Show less This action step requires the AI Skill Kit and appropriate roles, such as `actiondesigner`, to create and manage custom actions that include this step. ## Inputs You provide inputs to the step to control the skill execution: * **Capability:** Reference to a OneExtend Capability record, which can import capability inputs as action inputs. Note that some reserved words (e.g., `sysid`) cannot be used as action input names and require validation. * **Payload:** JSON-formatted string containing data sent to the Now Assist skill and OneExtend Framework for processing. * **Skill Config:** Reference to a Now Assist Skill Config record that defines the skill to be executed. * **If this step fails:** Choice to either continue running the next step or proceed to error evaluation. ## Outputs The step produces several outputs accessible in the Data panel to support subsequent action logic: * **Error:** Contains any error message from the skill execution; empty if successful. * **Provider:** Name of the Large Language Model (LLM) provider handling the request. * **Response:** Text output generated by the LLM, such as summaries or conversation replies. * **Result:** JSON-formatted string with detailed results from the LLM. * **Status:** Indicates the status of the call to the provider. * **Step Status:** Object containing runtime details of the step execution. * **Step Status Code:** Integer indicating success (0) or error (1) of the step. * **Step Status Message:** Contains system-generated error messages if any. ## Example: Calling the Flow Summarization Skill This example demonstrates invoking the Flow Summarization skill to generate a textual summary of a flow or subflow: * **Inputs:** Two action inputs map to skill inputs: `Flow to summarize` (Sys ID of the flow) and `Type` (such as flow, subflow, action). * **Capability:** Uses the Flow Summarization OneExtend Capability record. * **Payload:** JSON object mapping action inputs to skill inputs. * **Skill Config:** Uses the Flow Summarization Now Assist Skill Config record. The skill outputs: * No error output (indicating successful execution). * The LLM provider name (e.g., Now LLM provider). * A textual summary of the flow, which can be used as input for other actions or workflow logic. For example, the summary explains the Change - Normal - Authorize flow logic including approval checks and notifications. * A JSON result containing both the response and the flow elements used to generate it. * A success status code indicating the step ran without errors. ## Why it matters for ServiceNow customers This step simplifies embedding AI-driven capabilities into workflows by allowing seamless invocation of Now Assist skills. It enables automation designers to integrate AI-generated insights, summaries, or responses directly into workflow logic, enhancing decision-making and operational efficiency. With configurable inputs and outputs, customers can tailor AI interactions to their specific use cases and handle errors gracefully within their workflows. Run a published Now Assist skill from an action. Configure the Now Assist skill inputs and skill outputs from the step inputs and step outputs. ## Roles and availability {#call-now-assist-skill-step__section_q5n_hhn_w2c} Available as an Workflow Studio action step. Users with the action_designer role can create a custom action with one or more action steps.{#call-now-assist-skill-step__available-action} AI Skill Kit requirements{#call-now-assist-skill-step__action-ih-subscription} : This step requires the AI Skill Kit. For more information, see [Exploring AI Skill Kit](https://www.servicenow.com/docs/access?context=exploring-now-assist-skill-kit&version=yokohama&pubname=yokohama-intelligent-experiences&ft:locale=en-US). Role requirements{#call-now-assist-skill-step__action-fd-roles} : This step requires roles granted by delegated development or assigned to the user. For more information, see [User access to Workflow Studio flows](https://servicenow-prod.fluidtopics.net/GYXUUfIquOaOYZYuLzS2rQ "Administrators can grant users access to Workflow Studio flows by assigning delegated development permissions or directly assigning a user role. Administrators can also specify which features and content a user can access based on user roles. Application developers can access Workflow Studio functionality through APIs for flows, subflows, and actions."). ## Inputs {#call-now-assist-skill-step__section-inputs} Provide a value for each input that your action needs. To add dynamic values, you can also drag and drop pills from the Data panel or select them from the pill picker.{#call-now-assist-skill-step__inputs-description} Capability : Data type: Reference Reference to a OneExtend Capability \[sys_one_extend_capability\] record. When you select a capability record, the system gives you the option to import the capability inputs and create matching action inputs for them. Note: Some capability inputs use reserved words such as sys_id that can't be used as action input names. Review any inputs that are created by the import process and verify that the action input names are valid. Payload : Data type: String JSON-formatted string containing the payload to send to the Now Assist skill and the OneExtend Framework for processing. Skill Config : Data type: Reference Reference to a Now Assist Skill Config \[sn_nowassist_skill_config\] record. If this step fails : Data type: Choice Option to continue running the next step or go to error evaluation. To use the step status code or message for a custom action error condition, see [Action error evaluation](https://servicenow-prod.fluidtopics.net/CcC8uUbuR0i67ZKb17xIDA "Enable actions to catch step failures and continue running. Identify when specific error conditions occur and return your own action status code, status message, and error state."). ## Outputs {#call-now-assist-skill-step__section-outputs} These outputs appear in the Data panel. You can use them as inputs elsewhere in your action.{#call-now-assist-skill-step__outputs-description} Error : Data type: String Error message produced by the skill. This output is empty when the skill successfully runs the Now Assist skill request. Provider : Name of the LLM provider that fulfilled the Now Assist skill request. Response : Text response generated by the LLM such as a summary or a reply to a conversation. Result : JSON-formatted string containing the results generated by the LLM to the input payload. Status : Status of the call to the provider. Step Status : Data type: Object Object data pill containing runtime details about the step. Each step in an action returns a Step Status. Step Status Code : Data type: Integer Integer data pill indicating whether the step produced an error. A step returns a value of 1 when it produces an error for any reason. For example, a step can produce an error if it is missing mandatory input data or returns output in the wrong data type. A step returns a value of 0 when it runs successfully. You cannot customize these codes. * Success \[0\]: The step succeeded. * Error \[1\]: The step produced an error. {#call-now-assist-skill-step__ul_obs_2r1_j1c} Step Status Message : Data type: String String data pill containing the error message produced by the step or system operation. You cannot customize the step status message. ## Call the Flow Summarization skill {#call-now-assist-skill-step__example_hpq_5ln_w2c} This example action calls the Flow Summarize skill, which provides a text summary of what a flow or subflow does. The Flow Summarization skill has two skill inputs. * sys_id. This input contains the Sys ID of the component to be summarized. * type. This input contains the type of component to summarize. {#call-now-assist-skill-step__ul_rvm_q3m_y2c} To support the two skill inputs, this example action has two matching action inputs. * Flow to summarize. This input contains the Sys ID of the component to be summarized. * Type. This input contains the type of component to summarize such as flow, subflow, or action. {#call-now-assist-skill-step__ul_vsn_x3m_y2c} Note: An action can't have an input with the name sys_id, because that name is a reserved word. We can test this action by providing these example input values. Capability : This example uses the Flow Summarization OneExtend Capability record. Payload : This example uses a JSON object to map action input values to skill values. In this example, we provide the Sys ID of the Change - Normal - Authorize flow and the type of flow. Skill Config : This example uses the Flow Summarization Now Assist Skill Config record. Here are the example output values produced by the Flow summarization skill. Error output : This example doesn't produce an error output. Provider output : This example uses the Now LLM provider. Response output : This example uses the Flow Summarization skill to generate a summary for the Change - Normal - Authorize flow. Typically, you would use this Response output as an input for another action or flow logic. This flow, named "Change - Normal - Authorize", is designed to process a Change Request that uses the Normal Change Model and is in the state of Authorize, while ensuring it is not on hold. The flow begins by checking if the assignment group manager has already approved the change request. If not, it applies a Change Approval Policy, specifically for risk approvals. The flow then evaluates whether the approval process was skipped, rejected, or approved. If skipped, it does nothing. If rejected, it sends an email notification to relevant parties. If approved, it updates the change request record to mark it as approved. In parallel, the flow waits for the change request to be on hold or inactive, and then disregards any change request approvals. It also evaluates the change model again. The flow ends after completing these tasks. Result output : This example returns a JSON object containing the both the response and the flow elements used to generate the response. Status output : This example returns a success output value.