Skip to main content

Configure Genesys Agent Copilot settings

Series: Create a Genesys Agent Copilot

Previous suggested step: Create a new Genesys Agent Copilot

Next suggested step: Configure Genesys Agent Copilot natural language understanding

Settings tab

The Agent Copilot setup procedure navigates you to Settings.

General settings

  • You can rename the Agent Copilot.
  • The language displays the language that you initially selected, which you cannot change.

Context settings

Context definition

Click Define context to determine what context Agent Copilot uses to filter the knowledge it displays to the agent. The Context window appears. This window lists all the contexts for this Agent Copilot. In the Actions column, you can click Edit to edit a context, or Delete to delete a context. For more information, see .

To add a new context, click + New context.

  1. Enter a unique name for the context.
    Note: If you change a context name that is used in article filtering, you must delete and recreate the context article filter.
  2. Select a source type from the list.
    Note: The default source type is conversation attribute.
  3. Select an attribute schema from the list. For more information, see .
  4. After you select an attribute schema, select an attribute from the Attribute name list.

Knowledge settings (optional)

In the Knowledge content selection section, choose either Knowledge Base or Knowledge Configuration. If you select Knowledge Base, you can use a knowledge base you previously created in knowledge workbench V2. If you select a knowledge configuration, you can reference a knowledge fabric. For more information, see and , respectively.

Knowledge Base

From the Knowledge base selection list, select the knowledge base that you want your Agent Copilot to use.

You can filter the knowledge base to show only applicable articles to agents. To do this, click Filter. From the list, select Categories, Labels, or Contexts.

For more information on labels and categories, see

To use contexts for filtering, select a context by typing its name in the search box, then selecting it in the checklist. Choose whether articles must match all labels from the contexts, or whether they must only match any label. If you select match all, Agent Copilot does not search the knowledge base until all context values, such as the name of the active intent and the custom attribute value, are filled, for which it matches labels. If you select match any, Agent Copilot surfaces knowledge articles if at least one context value is filled, for which it matches a label.

If you apply more than one filter, the following rules apply:

  • If you select categories and labels, an AND relationship exists between categories and between labels; however an OR relationship exists with the categories, within the labels, and also between categories and between labels. For example: (Category 1 OR Category 2) AND (Label 1 OR Label 2).
  • If you select only  labels, an OR relationship exists between the labels
  • If you select only  category, an OR relationship exists between the categories

Knowledge configuration

From the Knowledge configuration selection list, select the knowledge configuration that you want your Agent Copilot to use. For more information, see .

To check which knowledge configuration settings correspond to the selected item in the list, click the arrow next to the Knowledge Configuration list.

 Additional settings

Knowledge confidence threshold

This setting allows you to adjust the sensitivity of articles being retrieved from the knowledge base for queries.

Note: Use caution when you adjust this threshold. A threshold that is too low can increase true positives but also increase false positives. A threshold that is too high can reduce false positives but also reduce true positives. Genesys recommends these default values:
  • 0.5 for non-content search knowledge bases
  • 0.65 for English content search knowledge bases
  • 0.7 for non-English content search knowledge bases
Additional guidance for content search knowledge bases

For content search knowledge bases in Filipino, Greek, Hebrew, and Malay, values above 0.7 may improve search quality. Genesys does not currently recommend a different default threshold for these languages. However, if your knowledge base uses one of these languages, consider testing higher threshold values to determine the optimal balance between search precision and recall for your content.

Tips:

  • If you use AI-generated answers, set the confidence threshold to 0.65 or below for all languages.
  • If a knowledge article supersedes an intent-based rule, lower the natural language understanding (NLU) confidence threshold.
  • Use phrases sparingly and evenly distributed across knowledge articles.
  • If you use a content search knowledge base, remove all phrasings. Make articles only title and body of contents.
  • If you use a content search knowledge base and the title is not long enough to summarize all the content of the article appropriately, add up to three to five phrases to fill any gaps.
Auto suggestions

Genesys Cloud Copilot automatically displays an answer compiled from the assigned knowledge base. 

Toggle Detect knowledge questions to allow the Agent Copilot to identify whether the query requires a knowledge search. This setting reduces the number of potentially irrelevant suggestions from Agent Copilot. Detecting knowledge questions is enabled by default for new Agent Copilots.

If Detect knowledge questions is enabled, you can toggle Refine search queries to allow Agent Copilot to reformulate the agent query. If you enable this, Agent Copilot can use the conversation context to reformulate and optimize the query for the knowledge search. To agents, Agent Copilot displays the reformulated query above the knowledge search result.

You can choose from the following options:

  • Articles only: The Copilot displays up to three relevant knowledge articles.
  • Answer highlights: The Copilot displays the answer from within the relevant knowledge articles that pertain to the customer’s query.
    Tips:
    • Write full sentences as opposed to bullet points. 
    • Write grammatically sound sentences with proper punctuation.
  • Answer generation: The Copilot generates an answer based on the content of the relevant knowledge articles in the assigned knowledge base.
    Note: Answer generation is only available for Knowledge bases in which content-based search is enabled.

For best practices, see

When enabled and an agent searches for a keyword or phrase, the Copilot displays an AI-generated answer compiled from the knowledge articles listed in the search results. Choose from the following options. Manual search is only available for knowledge bases.

  • Answer highlights: The Copilot displays the answer from within the relevant knowledge articles that pertain to the customer’s query.
    Tips:
    • Write full sentences as opposed to bullet points. 
    • Write grammatically correct sentences with proper punctuation.
  • Answer generation: The Copilot generates an answer based on the content of the relevant knowledge articles.

Natural Language Understanding settings

Confidence threshold: Adjust how the confidence threshold works in your organization’s NLU. This threshold determines which intents a model predicts based on a specified utterance.

Example: If the threshold is set at 0.65, the model only predicts an intent for an utterance when its confidence score meets or exceeds 0.65. To help enhance the model’s accuracy in predicting the correct intents, you can optimize the threshold. A threshold set too low can lead to more false positives, as the model can predict intents that are not a true match. Conversely, a threshold set too high could exclude relevant intents from being predicted. The recommended and default threshold is 0.4.

After Call Work settings

Wrap up codes

Display AI-powered wrap up codes: When enabled, Agent Copilot displays AI-powered wrap up codes. Agent Copilot can suggest up to three predicted wrap-up codes when the agent begins after-call work.

To set up wrap-up code prediction, admins must add descriptions to the respective wrap-up code.

To add a wrap-up code with description, follow the steps in , and enter a description as well. Always assign the wrap-up code to the assigned queue.

Summarization

Display AI-powered summaries: When enabled, Agent Copilot displays AI-powered summarization. If you enable an Agent Copilot on a queue, conversation summarization automatically prepares summaries of conversations between agents and customers. Genesys Cloud generates summaries for all agents in the respective queue. Only agents with the role can view these summaries.

Notes:
  • If you do not want Agent Copilot to generate and store summaries, disable both Display AI powered summaries  and Display AI-powered wrap up codes.
  • If you enable suppression for interactive voice response (IVR), hold, or queue wait times, Agent Copilot does not generate transcripts for these segments. As a result, summaries do not include any content from these segments. Genesys recommends that you enable suppression during these segments to enhance customer privacy. For more information, see .

Summary configuration

Select an existing summary configuration, or create a new one.

  • To create a new summary configuration, click the open button to open AI Studio in a new tab, and then add your summary configuration.
  • To use an existing summary configuration, select it from the list.

For more information about summary configuration, see .

Summary on demand

Choose whether agents can generate a summary for an interaction at any point before the interaction ends.

Agents can only create one summary per interaction. If the agent generates an on-demand summary, Genesys Agent Copilot does not generate a summary at the end of the interaction.

In case the interaction is transferred, Genesys Agent Copilot transfers the on-demand summary for the previous segment of the interaction. It does not transfer any edits the agent made to the on-demand summaries in the previous segment of the interaction.

Auto opt-in

Choose whether Genesys Agent Copilot must always use the latest model to generate summaries. This configuration applies to both standard summaries and custom summaries. If you opt in and if there is a newer LLM model available, Genesys Agent Copilot uses it to generate summaries. If you opt out, the Genesys Agent uses the older model until the model becomes deprecated. You can also toggle to opt in or out of an existing Copilot.

Note: In a multilingual configuration, additional languages inherit the opt-in setting by default. You can change the setting for each additional language.

Summaries time to live

Specify how long Genesys Agent Copilot stores summaries. The value must be between one hour and 10 days. If you do not specify a time, Genesys Cloud stores summaries for 10 days.

Notes:
  • If the after call work timeout is longer than the summaries time to live, summaries only appear in the UI and changes or edits are not saved.
  • If you add multiple languages to your Copilot, the time to live is the same for each new language as the original language by default. You can override the time to live for each language.
  • If you change the time to live setting for a Copilot, the new value only applies to summaries generated after the change.

Click Save Copilot.

Tip: Admins can use the API or copy and paste interaction summaries into CRM or interaction notes.