--- sourceDocument: Zurich Enable AI sourceDocumentLink: https://www.servicenow.com/docs/r/zurich/intelligent-experiences Release : - zurich ft:locale : - en-US ft:publication_title : - Zurich Enable AI ft:clusterId : - platai bundleId : - platai workflow : - Platform --- # Create an AI voice agent # Create an AI voice agent {#ariaid-title1} * Release version: Zurich * * Updated March 12, 2026 * * ![](https://www.servicenow.com/docs/portal-asset/ico-clock) 8 minutes to read Create an AI voice agent in the AI Agent Studio to resolve cases, incidents, or tasks through the phone channel. ## Before you begin Role required: sn_voice_aia.admin ## Procedure 1. Navigate to the New AI Agent page in the Assistant Designer or AI Agent Studio. {#create-a-voice-enabled-ai-agent__table_emg_jm5_w2c__entry__2} | To | Do this | |-|-| | Navigate in Assistant Designer | 1. Go to the Asset Library and select Create asset. The Create asset window appears. 2. Select AI Agent. You are redirected to the New AI Agent workflow in the AI Agent Studio. {#create-a-voice-enabled-ai-agent__ol_ykb_qm5_w2c} | | Navigate in AI Agent Studio | 1. Go to AllAI Agent StudioCreate and manage and select the AI agents tab. 2. In the Add drop-down list, select Voice to create an AI voice agent. | [ ] {#create-a-voice-enabled-ai-agent__table_emg_jm5_w2c} 2. On the Define the specialty page, describe your AI agent and provide instructions on how you want your AI agent to perform its tasks. Select Generate details to generate a description and instructions with AI. If you provide the description of what you want the agent to do, you can select Generate to write the name, description, AI agent role, and instructions fields for you. You can change those fields after the text has been generated or try again with new instructions. Note: The more details that you provide, the more accurately your AI agent can perform. 1. On the form, fill in the fields. {#create-a-voice-enabled-ai-agent__table_mp3_4nj_zcc__entry__2} | Field | Description | |-|-| | AI agent name | Name of the AI agent. Provide a name according to the outcome that you want your AI agent to achieve. For example: Incident manager. | | AI agent description | Brief summary of what your AI agent can do autonomously. Outline all the activities that you want your AI agent to perform. Example for the Incident manager: The Incident manager AI agent assists users with existing incidents by gathering information over the phone, matching it to records in the system, and sharing relevant incident details. It can also make limited updates to the incident using preconfigured tools. | [Table 1. Give a unique name and description form] {#create-a-voice-enabled-ai-agent__table_mp3_4nj_zcc} 2. On the form, fill in the fields. Note: The AI agent uses this information as guidance to tailor its responses and actions. {#create-a-voice-enabled-ai-agent__table_uj3_msj_zcc__entry__2} | Field | Description | |-|-| | AI Agent role | Capabilities and responsibilities for your AI agent. Roles enable your AI agent to perform its required actions. Example for the Incident manager: You're an AI agent that communicates incident details and makes changes to incidents for callers. You can retrieve incident fields and comments, and escalate urgency. Only gather information and run tools if the caller matches the user listed on the incident. | | List of steps | Necessary steps to be followed by the AI agent while carrying out its role. Note: The instructions are sent to the large language model (LLM). Example for the Incident manager: Incidents are records for tracking issues and their resolution updates through a set of journal entries. Your objective is to act as an incident manager. If the user provides an incident number, verify the caller identity before sharing permitted details. If no number is given, match incidents using the short description and confirm with the user. Allow updates only if the user matches the caller. Offer live agent transfer if the user is unsatisfied. To resolve an incident: 1. Verify the caller number. 2. Search for incident records associated with the user. 3. Consider the context and relationship between different data points. 4. If the user provides an incident number, verify their identity against the caller field before sharing permitted details. Proceed with updates or modifications only after the caller's identity is verified. 5. If no incident number is given, search using the short description, list matching incidents with brief summaries, and confirm which one the users wants to explore further. 6. Only share allowed fields (for example, incident number, caller, dates, urgency, state, short description, comments). Summarize comments if there are more than five, and notify the user if a field is empty or restricted. 7. If the user is unsatisfied, offer to transfer to a live agent. End the conversation politely if they're satisfied or after a transfer. {#create-a-voice-enabled-ai-agent__ol_z5v_lwj_d2c} | [Table 2. Define the role and required steps form] {#create-a-voice-enabled-ai-agent__table_uj3_msj_zcc} Follow these guidelines for writing effective Instructions: * Define the AI agent's role: * Clearly state the primary function of the AI agent in one or two sentences. * Example: The AI agent acts as a customer service assistant. {#create-a-voice-enabled-ai-agent__ul_ly5_c3c_fdc} * Outline specialties: * Specify the key areas or tasks that the AI agent handles. * Example: Specializes in handling inquiries and resolving customer issues. {#create-a-voice-enabled-ai-agent__ul_t3g_m3c_fdc} * Identify the business problem: * Articulate the specific business challenge for the AI agent to address. * Example: Reducing customer wait times by 50%. {#create-a-voice-enabled-ai-agent__ul_qjk_p3c_fdc} * Describe the workflow: * Provide a brief scenario of how the AI agent is to be used. * Example: Automating responses to common queries and escalating complex issues to human agents. {#create-a-voice-enabled-ai-agent__ul_jpd_53c_fdc} * Be concise and clear: * Use simple and direct language. * Avoid jargon and technical terms. * Example: The AI agent helps customers find answers quickly. {#create-a-voice-enabled-ai-agent__ul_prk_z3c_fdc} * Provide actionable steps: * Include specific instructions on how to set up and use the AI agent. * Example: Step 1: Define the types of inquiries the AI agent handles. Step 2: Integrate the AI agent with the customer service platform. {#create-a-voice-enabled-ai-agent__ul_d2y_p4c_fdc} * Include examples: * Provide real-world examples to illustrate how the AI agent should function. * Example: For example, the AI agent can automatically respond to questions about the order status. {#create-a-voice-enabled-ai-agent__ul_t4x_npc_fdc} * Focus on outcomes: * Emphasize the benefits and outcomes of using the AI agent. * Example: Using the AI agent leads to faster resolution times and higher customer satisfaction scores. {#create-a-voice-enabled-ai-agent__ul_w4x_npc_fdc} {#create-a-voice-enabled-ai-agent__ul_ohy_zhc_fdc} By following these general guidelines, you can create clear and effective prompt instructions that enable you to use AI agents to their fullest potential. For more information and examples, see [General guidelines for creating AI agents](https://www.servicenow.com/docs/~MlCy7RCcV46CgQnJ0PsGw "By following some general guidelines for creating AI agents and agentic workflows, you can create clear and effective instructions that help maximize their efficiency and effectiveness."). 3. **Optional:** Determine if the AI agent can be accessed by third parties. Making your AI agent discoverable allows you to use your ServiceNow AI agent on other platforms. You can still use it on your ServiceNow instances as well. 4. Select Save and continue to move to the next step. {#create-a-voice-enabled-ai-agent__substeps_c1m_4m3_d2c} 3. In the Add tools and information tab, configure additional tools and data sources that provide capabilities necessary to your AI agent to accomplish its objectives. Select Recommend Tools for AI to suggest the tools that are required for the AI agent to carry out the tasks that it's built for. AI suggests tools based on the AI agent description and instructions given in the previous section. You can add the tools suggested by AI or manually select the tools from the Add tool drop-down list. You must add at least one tool to continue setting up your AI agent, but you can also add more tools to your AI agent. The data input and output type for tools must be string for optimal voice experience. The following tools are available for AI voice agents: * [File upload](https://www.servicenow.com/docs/YsSnjFb0Ma5gAL6E_8lSWA "Upload files for analysis by an AI agent in AI Agent Studio to grant your AI agent access to specialized knowledge."): Different file types such as PDF, DOCX, or TXT formats that you can add to your AI agent. * [Flow action](https://www.servicenow.com/docs/99lsO4BoLMcZzlr5_j1~Ag "Add a flow action to an AI agent in AI Agent Studio. Define the flow action to use it as a reusable operation in automating the ServiceNow AI Platform features without having to write code."): Custom automated processes in your system that you can add to your AI agent. Example for the Incident manager agent: Fetch details of the incident. * [Knowledge Graph](https://www.servicenow.com/docs/9KMK5xoF6niaiB_XcVJHkg "Add a Knowledge Graph to an AI agent in AI Agent Studio that uses the structured and unstructured data from different ServiceNow records to enhance the performance of AI agents."): Various Knowledge Graph items that you can add to you AI agent. * [MCP server tool](https://www.servicenow.com/docs/TLuo5WqrdBc_F84OInP8SQ "Add a Model Context Protocol (MCP) tool to an AI agent in the AI Agent Studio so your users can access the MCP server."): An MCP server tool that you can to your AI agent. * [Record operation](https://www.servicenow.com/docs/Vqnn0khtNvCX2fFe5aMg_Q "Add a record operation to an AI agent in AI Agent Studio to create, update, look up, or delete records."): Different record operations that you can add to your AI agent. * [Script](https://www.servicenow.com/docs/8hD08A9BN7_jM3UIxazWkg "Create a script to add it to an AI agent in AI Agent Studio. With scripts, you can use the scriptable APIs and back-end integration to support the AI agent."): Editable scripts and APIs that you can add to your AI agent. * [Search retrieval](https://www.servicenow.com/docs/cviO60xWUi~mfirhn9rWPg "Add a search retrieval to an AI agent in AI Agent Studio. Leveraging the Retrieval-Augmented Generation (RAG) enables an AI agent to retrieve and incorporate relevant information from an external source."): Information retrieval processes in your system that you can add to your AI agent. Note: Create a dedicated search profile that includes only the KB articles for AI voice agents to reduce the search scope and minimize latency. * [Sub flow](https://www.servicenow.com/docs/8cU22AK7ZKU4G1hxlFQHOQ "Add a subflow to an AI agent in AI Agent Studio. Subflows are reusable sequences of processing steps that can be called from within a flow."): Automated flows in your system that you can add to your AI agent. {#create-a-voice-enabled-ai-agent__ul_zbv_1hp_hgc} 4. In the Define security controls tab, define who can access the AI agent and what data the AI agent has access to. 1. Define the users who can access this AI agent (ACLs). Choose from: * Any authenticated user: A user who is logged in can access the AI agent. * Users with specific roles: Users that are assigned specific roles can access the AI agent. Selecting this option enables you to select the roles in the Role drop-down list. * Public: Anyone can access the AI agent, even without logging in. {#create-a-voice-enabled-ai-agent__ul_s4d_42p_hgc} 2. **Optional:** If you selected Public, select Require caller identification to have the AI voice agent identify the caller before the conversation begins. When selected, the agent uses the caller identification methods configured in the Caller Identification section of the voice assistant to match the caller to a name. Identification does not authenticate the caller or grant additional access. The session continues with public user permissions. Note: This option is available only for voice agents configured with Public access. 3. Define the user identity of the AI agents to determine what data it has access to. The default selection is Dynamic user. The user passes their roles to the AI agent, and the AI agent runs as the user that invokes it. The user's ACLs determine the data accessible to the AI agent. {#create-a-voice-enabled-ai-agent__substeps_o4j_yqw_nhc} 5. Select channels and activation status for your AI voice agent. 1. Select Allow to enable users to use phone calls to invoke the AI agent. 2. In the Choose voice assistants field, select the default assistant or a custom assistant. Voice assistants are created in Assistant Designer. See [Create a voice assistant](https://www.servicenow.com/docs/access?context=configure-voice-assistants&version=zurich&pubname=zurich-conversational-interfaces&ft:locale=en-US) for more information. 3. Activate the AI agent by toggling the Status. 4. **Optional:** Select Test in assistant to test your AI voice agent in the voice testing experience. * If one voice assistant is linked, the voice testing panel opens directly. * If multiple voice assistants are linked, select the voice assistant you want to use for testing. * If no voice assistants are linked, Test in assistant is disabled. Link a voice assistant first. {#create-a-voice-enabled-ai-agent__ul_test_in_assistant_onh} {#create-a-voice-enabled-ai-agent__substeps_onh_jzx_dfc} 6. Select Done to save the configuration. *[\>]: and then