Define an agent’s role and guidelines

AI-generated summary

Genesys Cloud's Virtual Agent platform integrates AI-powered conversation flows through AI Studio, featuring comprehensive agent configuration and management capabilities. The system utilizes a guide creation system with core commands (Say, Ask, Ask and Confirm, Call) and supports multiple languages while maintaining English command syntax. Command implementation requires specific formatting conventions and enables variable integration, data actions, and flow control through If:Then:Else logic and navigation commands. The platform's agentic virtual agents require specific permissions and are designed to be domain-specific with defined roles and personalities. Through the Agent profile interface's five main sections (Agent, Tools, Knowledge, Configuration, and Guardrails), organizations can configure detailed agent behaviors, data collection processes, and conversation management rules. The system supports voice interactions through ASR transcriptions and handles multi-step conversations with specific formatting requirements. Implementation best practices emphasize single-task focus, proper variable management, and comprehensive validation through the guide checker. Integration with Architect and digital bot flows occurs via Call Guide action, with automatic updates for Virtual Agent-enabled flows. The platform ensures privacy policy compliance, provides community support, and requires complete configuration of all agent sections before publishing, though work-in-progress configurations can be saved during setup.

Series: Create an agentic virtual agent

Previous suggested step: Create an agentic virtual agent

Next suggested step: Set up tools for an agentic virtual agent

Feature coming soon: Agentic virtual agents

PrerequisitesClick to expand

Establish the role, capabilities, and objectives of the agentic virtual agent and also outline the guidelines and best practices for its effective implementation and use. By design, a single agent is domain-specific and cannot handle all use cases or scenarios.

Define the role

Use the role to indicate the personality and branding that the agent represents. To confine the agent to its intended use, you can also specify what is outside its scope.

Click Menu > Orchestration > AI Studio > Agentic Virtual Agents.
(Optional) To search for a specific agent, start typing the name of the agent in the search box. As you type, the matching agents are shown.
To open an agent, click the agent’s tile. The Agent profile view opens.
Under Role, specify the agent’s intended personality, branding characteristics, and capabilities.

Agent role examplesClick to expand

Do / Don’t	Example	Explanation
	You are a banking assistant for ACME Bank that provides simple banking services via chat or voice.	Mentioning the role with the brand name and a high-level explanation in the first sentence. Afterward, you can highlight the capabilities.
	You are a banking assistant for ACME Bank who helps customers with their loan application queries. When beginning a conversation with a customer, you always maintain a professional and polite tone.	Being clear and concise.
	You have the ability to query a knowledge base to provide users with more information about ACME personal bank loans. You have the ability to register a user for a personal loan by collecting their details step by step.	Ensuring that you include a high-level description of the available tools and capabilities.
	You have the ability to help a user apply only for a personal loan of up to ten thousand dollars.	Also mentioning what may be within the domain but out of scope for the agent.
	You are a helpful assistant.	Telling it to be a ‘helpful assistant’ is a bad example, as it may try to be helpful and go outside the intended scope.
	You handle queries from a customer.	Not providing any information about what it needs to do.
	You help customers with queries related to the travel company.	Providing ambiguous and vague instructions.
	Testing agent xyz	Using words such as ‘test’ or ‘testing’.

Set guidelines

Outline some instructions to the agent for the entire context of the conversation. This is the set of guidelines that the agent must follow.

Under Guidelines in the Agent tab, click Add. An empty text field appears.
Write an instruction that guides the agent in its conversations. For example, “If a question is ambiguous, ask for clarification.”
Add as many guidelines as needed to guide the agent throughout an interaction session.
Click Save.

Guideline examplesClick to expand

Do / Don’t	Example	Explanation
	Always validate users before providing balance data. Include the currency. If a question is ambiguous, ask for clarification. If asked about services, features, or assistance beyond your available functions, state that you cannot help with those requests.	Provide positive examples of what it can do. Being positive helps the LLM focus on what it is trying to be.
	You are a voice agent (in voice mode) that receives transcriptions of the user’s speech from an ASR system. If you receive partial inputs, do not apologize or say that you didn’t hear them. Instead, confirm what you understood from the input, and if it seems the user has paused, ask them to continue when ready. Example: User: My login code is 2 4 5 Assistant: So far, I heard 2 4 5. What are the remaining five digits? User: 2 4 5 6 6 5 7 8 Assistant: I heard your code as 2 4 5 6 6 5 7 8. Did I hear that correctly?	Specify here that you are on voice and how you want the agent to behave. Providing examples is always recommended.
	If a question is ambiguous, ask for clarification.	General instructions on how to handle situations when things don’t make sense.
	When specifying a range such as 1–3 or 10–20, never use a hyphen. Always use the word “to” instead (for example, 1 to 3 or 10 to 20).	Explain nuances like these when they help the voice TTS system.
	Avoid repeating the exact same question. If a user doesn’t respond as expected, rephrase the question, provide examples, or approach the topic differently.	Provide guidance on handling conversations where there may be confusion.
	You don’t need to explain every single tool.	List all capabilities in the instructions.
	Get Personal Data: – Ask “What is your name?” – Store in {{Variable.name}} – Ask “What is your age?”	This is the format that you can use in AI Guides. However, this format is not intended for agentic virtual agents.
	Always respond in the language of the user’s conversation	The model is already smart enough to do this.

Navigating the agentic virtual agent profile

The profile contains the following sections and a preview widget:

Show meClick to expand

You can switch to any of the sections of the agentic virtual agent while creating a new agent, editing an existing agent, or viewing an agent. Any incomplete section is indicated by the icon preceding the section name. You can save an agent with an incomplete section but you cannot publish the agent until all sections are complete.

After you have completed all necessary sections of the agent setup, you can click Publish to publish the flow.

[NEXT] Was this article helpful?

Get user feedback about articles.

Define an agent’s role and guidelines View summary