Connect an agentic virtual agent to knowledge

AI-generated summary

Genesys Cloud enables agentic virtual agents to connect to knowledge sources through the Knowledge Aggregate feature, allowing agents to access relevant data and context during conversations. Users configure these connections by navigating to Menu > Orchestration > AI Studio > Agentic Virtual Agents, selecting an agent, and accessing the Knowledge tab to enter a meaningful name, description, and select from available knowledge configurations or bases. The platform supports multiple languages including English (US, GB, Australia), Spanish, French, German, Portuguese, Japanese, Korean, Hindi, Arabic, Dutch, Italian, Polish, and Thai, with varying support levels depending on the knowledge resource type. Configuration requires Virtual Agent enablement in the organization, appropriate AI Studio permissions, and full Knowledge and Analytics permissions. Best practices for naming knowledge connections include domain-specific terminology that clarifies purpose (for example, "search_banking_faq_knowledge_base"), while descriptions should explain the knowledge base type and content at a high level. Pre-instructions configured in the Input tab establish conditions that must be met before agent access, including comprehensive descriptions of knowledge base contents and explicit statements of what information is not included to prevent agent confusion. Instructions must specify operational requirements such as translating queries into English and formatting them as questions before execution. Output delivery is configured through Outcome Instructions in the Output tab, where administrators define no-result response messages for unmatched queries and follow-up messages after successful matches. These instructions serve as recommendations that agents paraphrase contextually rather than deliver verbatim. The tab provides consolidated review of all configuration parameters before creation, allowing administrators to verify alignment between knowledge topics and the agent's defined role and guidelines. The agentic virtual agent profile integrates multiple sections—Agent, Tools, Knowledge, A2A, Configuration, and Guardrails—accessible during agent creation, editing, or viewing. Agents can be saved with incomplete sections, but publication requires all sections to be fully completed before the Publish button can be activated. Performance may vary across languages, and Genesys recommends testing with relevant languages before deployment. Related documentation covers agentic virtual agents, AI Studio, knowledge fabric, Genesys Virtual Agent, A2A connection configuration, agent preview functionality, and testing and troubleshooting procedures, with a structured feedback mechanism available for documentation improvement

Series: Create an agentic virtual agent

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

PrerequisitesClick to expand

Knowledge supplies relevant data sources and more context for your agentic virtual agents during conversations.

Connect to knowledge

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 AI agent, click the agent’s tile. The Agent profile view opens.
Click the Knowledge tab.
In the Name field, enter a meaningful name. For example, “search_banking_faq_knowledge_base.”
In the Description field, enter a high-level summary about the knowledge connection.

In the Knowledge section, select a knowledge configuration or base from the Choose your knowledge connection list. A preview tile of the selected knowledge configuration or base appears showing its connected sources and sync information. For more information about how to create a knowledge configuration, see . Similarly, for more information about how to add a knowledge base, see .

Note: Agentic virtual agent performance may vary across languages. Genesys encourages testing with languages relevant to your use cases before you go live.

Supported languages for knowledge configuration and baseClick to expand

Language	Region	Language code	Knowledge configuration	Knowledge base
Arabic	United Arab Emirates	ar-AE	✓	✓
Dutch	Belgium	nl-BE	✓
Dutch	Netherlands	nl-NL	✓	✓
English	Australia	en-AU	✓	✓
	Great Britain	en-GB	✓	✓
	United States	en-US	✓	✓
French	Belgium	fr-BE	✓
	Canada	fr-CA	✓	✓
	France	fr-FR	✓	✓
German	Germany	de-DE	✓	✓
German	Switzerland	de-CH	✓	✓
Hindi	India	hi-IN	✓	✓
Italian	Italy	it-IT	✓
Japanese	Japan	ja-JP	✓	✓
Korean	Korea	ko-KR	✓	✓
Polish	Poland	pl-PL	✓
Portuguese	Brazil	pt-BR	✓	✓
Spanish	United States	es-US	✓	✓
Spanish	Spain	es-ES	✓	✓
Thai†	Thailand	th-TH	✓

Click Next to proceed to the Input tab.

Examples of knowledge connection name and description

Knowledge connection name examplesClick to expand

Do / Don’t	Example	Explanation
	search_banking_faq_knowledge_base	Explains that it is a knowledge base but also that it is a banking FAQ source.
	search_mortgage_and_general_loan_knowledge_base	Informs the model that the knowledge base contains information about mortgage and general loan questions.
	search payment and transaction knowledge base	Domain-specific information about what the knowledge base covers. The model can understand its purpose and can differentiate it from other tools.
	search_knowledge	Too vague.
	knowledge tool that gets information	Vague and not domain-specific. Does not give any relevant information.

Knowledge connection description examplesClick to expand

Do / Don’t	Example	Explanation
	Issues the search query against a knowledge base with answers to consumer banking FAQs.	Explains what kind of knowledge base that it is at a high level.
	A knowledge search tool.	Does not tell the agent when to call it and how to use it. It is too vague.

Set up input conditions

Add pre-instructions

Under Pre-instructions in the Input tab, click Add and add any conditions that must be met before this knowledge connection is used by the agent. For example, “The knowledge connection contains banking FAQs covering account management, online banking, and mobile banking.”
Add as many pre-instructions as needed for the intended knowledge usage.

Knowledge pre-instruction examplesClick to expand

Good pre-instructions answer these questions:

Why should the agent call the knowledge configuration?
When should the agent call the knowledge configuration?
What should the agent do before it calls the knowledge configuration?

Do / Don’t	Example	Explanation
	The knowledge base contains banking FAQs that cover account management, online banking, mobile banking, security and fraud, loans and credit, deposits and withdrawals, fees and charges, cards and payments, investment services, and customer care.	Gives more context about what is in the knowledge base.
	If the user has trouble authenticating (for instance, if they cannot remember their password or email), or if their accounts are not as expected, do not call the FAQ, as it does not contain information about those issues.	Includes what is not in the knowledge base. Especially for cases where the agent gets confused. For example, mortgage loan information only.
	Always translate the query argument into English before using this search function. The query must be in the form of a question, for example: “How long does it take for transfers between accounts to settle?” There is no way to call this function with the query argument in a different language.	Specific instructions that tell the agent what to do beforehand (for example, if you know that the knowledge is in a different language, you can adjust the data sent by the model).
	This knowledge tool has all the information about the banking procedures.	Too vague and high level.

Define required variables

Link the existing variables that must be collected before the knowledge connection is actioned.

To mark a variable as required, follow these steps:

Under Required variables in the Input tab, click Add. The Required variables dialog box opens.
In the Name field, enter a meaningful name for the variable linkage.
Under Link variable, click Select. The Link existing dialog box opens.
Select a variable from one of the following tabs:
- Tool Input
- Tool Output
- Start context
Click Select to confirm the variable selection.
Click Add.

These linked variables are now available for further use in the validation rules which you can configure in the Validation checks section.

Add validation checks

Define strict validation rules that the agent must pass before the tool is actioned. The system offers two flexible ways to create your preconditions:

Builder: Use the guided interface to build the validation logic to meet your requirements. This mode is recommended for common rules and regular format checks.
Code: Use the code editor to write custom logic for advanced or unsupported validation checks.

Considerations

In Builder mode, you can add up to five groups. Each group can contain up to five checks. Also, Builder mode does not support nested grouping.
Builder mode supports standard conditions. To work with arrays or create more complex groupings, use Code mode.

To create an out, follow these steps:

Under Validation checks in the Input tab, click Add and add any validations that must pass before this tool is used by the agent.
Select either the Building or Code mode for building your validation logic:

If you chose the Builder mode, follow these steps:

Use the toggle switch to select between the AND and OR operator.
From the Select input list, select one of the required input variables.

From the Select condition list, select the required condition.

Condition	Description
is null	Returns true if the value is empty or not set.
is not null	Returns true if the value is present.
equals	Returns true if the value exactly matches the specified value.
does not equal	Returns true if the value does not exactly match the specified value.
in	Returns true if the value matches any value in the specified list.
not in	Returns true if the value does not match any value in the specified list.
contains	Returns true if the value includes the specified text.
doesn't contain	Returns true if the value does not include the specified text.
begins with	Returns true if the value starts with the specified text.
ends with	Returns true if the value ends with the specified text.
less than	Returns true if the value is less than the specified value.
less than or equal to	Returns true if the value is less than or equal to the specified value.
greater than	Returns true if the value is greater than the specified value.
greater than or equal to	Returns true if the value is greater than or equal to the specified value.

In the Enter value field, enter the value that must match against the selected condition.
Click Add check to more validation checks.
Click Add group to add a separate group of checks along with its own conditional operator.

If you chose the Code mode, enter your custom validation logic.
In the Failure instruction field, enter the instruction that the agent must follow when the check is not met. Include the next action, such as asking the participant to correct or provide the required information, or end the conversation if necessary.
Click Create to add the validation check.

Edit or delete validation checks

Under Validation checks in the Input tab, hover over the validation check that you want to edit or delete, and then click Edit or Delete.

Reorder validation checks

Under Validation checks in the Input tab, hover over the validation check that you want to reorder, and then click More.
Select the appropriate option to move the validation check up or down in the list. You can also add a new validation check above or below the current one.

After defining the required variables and adding validation checks, click Next and proceed to the Output tab.

Set up output delivery

Under Outcome Instructions in the Output tab, add the following response messages:
- No result response: The message (or a similar message) that the agent sends when the agent finds no matching result.
- Follow-up response: The message (or a similar message) that the agent sends as a follow-up after sending the matched result.
Click Next and proceed to the Summary tab.

Note: The outcome instructions are not mandatory requirements but recommendations for the agent. They are not delivered verbatim but are paraphrased based on the conversational context.

Review and save the configuration

Get a consolidated view of the entire configuration, identify any improvement areas, and tweak the settings before saving the configuration.

In the Summary tab, review all the parameters of the knowledge connection.
To edit the parameters, click Edit.
Click Create.

Note: Ensure that the topics and questions this knowledge configuration contains are covered in the description (scope) of the agent’s role and guidelines.

Navigate 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.