Rules

The Rules engine within AI Studio allows you to define custom rules that monitor live customer interactions for specific keywords and phrases. When these keywords or phrases are detected, the engine automatically triggers helpful actions or suggestions to assist agents in real time.

The engine is powered by the Realtime Insights Orchestrator (RTI), which monitors live interactions, detects defined keywords, phrases, or conditions, triggers relevant actions or suggestions, and delivers them to the agent through the Copilot interface.

The Rules engine offers several key benefits that enhance its usability and effectiveness. It is customizable, allowing tenants to define their own rules and actions based on specific business needs. The engine is highly flexible, supporting a wide range of triggers and responses to suit various interaction scenarios. It is also integrated seamlessly with other Copilot for Agents features such as Real-time Summary, Sentiment Analysis, and Automated Summary (AutoSummary), ensuring a unified experience. Lastly, the engine is scalable, built to adapt to evolving business requirements and diverse interaction types across different industries.

You can configure up to 100 rules, with a maximum of 10 actions per rule.

Manage Rules

  1. Click the app selector icon of app selector and select AI Studio > Rules. The application displays the Rules interface.

Create New Rule

  1. In the Rules interface, click Add Copilot Rule.

  2. In the General tab, do the following:

    1. In the Name field, enter a unique name for the rule.

      Rule names must be unique. Rule names must not exceed 250 characters.

    1. In the Description field, enter a brief explanation of what the rule does.

    2. In the Notes field, add any optional comments or context for future reference.

  3. In the Rule Logic tab, do the following,

    1. In the Observation Interval field, select the time range in the call transcript that the application should monitor for rule evaluation.

      This setting does not trigger the rule. It only defines when in the call the application should begin checking for the selected keywords or phrases. Choose one of the following options:

      • After – The application starts evaluating the transcript after a specified number of seconds from the start of the call. For example, check the transcript after 20 seconds until the end of the call.

      • Between –The application evaluates the transcript only within a defined start and end time (in seconds). For example, check only between 30 and 90 seconds.

      • None – Choose this if no time-based condition is required. The system evaluates the entire transcript.

      • Within the First – Specify a time frame (e.g., first 30 seconds) within which the rule should trigger.

        Enter the required values based on the selected interval type.

    2. In the Rule Triggers section, define the conditions that activate the rule. You can configure the rule using one of these trigger types:

      1. Keywords or Phrases

      2. Intent Based Conditions

    3. Select the Keywords Or Phrases option to activate a rule when the system detects specified text in the interaction transcript.

      1. In the Add Keywords or Phrases field, click Add Keywords to enter one or more keywords or phrases that the system should look for within the selected interval. Use Clear to remove previously added keywords or phrases.

      2. In the Apply Rule When field, specify the condition that determines how the rule should evaluate the presence or absence of the entered keywords or phrases. Select one of the following options from the dropdown:

        • Any keyword or phrase is present: The rule activates if at least one of the selected keyword or phrase appears in the interaction.

        • All keywords or phrases are present: The rule activates only if every keyword or phrase you selected appears in the interaction.

        • Any keyword or phrase is missing: The rule activates if at least one of the selected keywords or phrases is not found in the interaction.

        • All keywords or phrases are missing: The rule activates only if none of the selected keywords or phrases appear in the interaction.

      3. In the Party That Activates Rule field, specify whose speech the system should evaluate when looking for the keywords. Select one of the following options from the dropdown:

        • Customer: Evaluate only customer utterances.

        • Agent: Evaluate only agent utterances.

        • Either Party: Evaluate utterances from the agent or customer.

      4. In the Activation Threshold field, enter the number of times the rule’s condition must be met before the rule activates for the first time.
    4. Select the Intent Based Conditions option to activate a rule when the system identifies a specific customer intent. The application performs intent detection only on customer transcripts.

      1. In the Rule Intent field, enter the intent that activates the rule. The intent name supports up to 100 characters. You can optionally enter a description and example utterances. Each field supports up to 500 characters.

      2. Use the Confidence Threshold slider or numeric field to define the minimum confidence score required for the detected intent to activate the rule. The supported range is 0.0 to 1.0, and the default value is 0.7.

        The rule activates only when the detected intent confidence score is greater than or equal to the configured threshold.

    5. The Activation Limits section defines how frequently the rule can trigger and how long it must wait between activations during a single interaction. This section includes the following fields:

      • Occurrence Limit:The maximum number of times the rule can trigger within a single interaction. Once the rule has been triggered up to the Occurrence Limit, it will no longer be activated during that interaction.

      • Cool Off Period: The minimum amount of time, in seconds, that must pass before this rule can be triggered again.

    6. Click Save. The rule is created and you can link actions to these rules.

  4. In the Linked Actions tab, link the rule to an existing action. To link the rule to an existing action, do the following,

    1. In the Linked Actions list, select the check box next to each action you want to link, or you can use the Search Actions box to find a specific action.

    2. Click Save.

Activate a Rule

  1. Identify the rule you want to activate in the Rules list.

  2. Click the More options icon () next to the rule.

  3. From the menu, select Activate.

  4. The rule’s status changes to Active in the list.

Deactivate a Rule

  1. Identify the rule you want to deactivate in the Rules list.

  2. Check the Status column to confirm the rule is Active.

  3. Click the More options icon () next to the rule you want to deactivate.

  4. From the menu, select Deactivate.

  5. In the Are You Sure? dialog box, review the profiles that will be impacted. Click Deactivate to confirm.

  6. The rule’s status changes to Inactive in the list.

Edit a Rule

  1. Locate the rule you want to edit in the Rules list.

  2. Click the More options icon () next to the rule you want to edit.

  3. From the menu, select Edit.

  4. Make the necessary changes in the rule configuration screen.

  5. Click Save to apply your changes.

Filter Rules by Status

  1. On the Rules page, click the Filter icon next to the search box.

  2. Select one of the following options:

    • All to view all rules.

    • Active to view only active rules.

    • Inactive to view only inactive rules.

  3. The list updates to show rules based on your selection. The applied filter appears above the list.

  4. To remove the filter, click Clear All.

  1. In the Linked Actions tab, click on Link Action.

  2. In the Link Action dialog box, select an Action Type from the dropdown menu, for example, Text Card, Checklist, Decision Tree, and so on.

  3. Based on the selected type, additional fields may appear:

    • For Text Card, choose a Card Type.

    • For a Checklist, select a checklist from the Select Checklist dropdown.

    • For a Decision Tree, select a decision tree from the Select Decision Tree dropdown.

  4. Click Link Action to add the new action to the existing list of actions.

Edit an Action

  1. In the Linked Actions tab, locate the action you want to modify.

  2. Click the More options icon () next to the action name.

  3. Select Edit from the dropdown options.

  4. Update the action details such as Action Type and other related fields.

  5. Click Update Action to apply your changes.

Activate or Deactivate an Action

  1. In the Linked Actions tab, locate the action you want to modify.

  2. Click the More options icon () next to the action name.

  3. From the menu, select Activate.

  4. The action’s status will change to Active.

    Similarly, you can deactivate a action.

Remove an Action

  1. In the Linked Actions tab, locate the action you want to modify.

  2. Click the More options icon () next to the action name.

  3. From the menu, select Remove.

    The action is removed from the rule.

Bulk Remove Actions

  1. In the Linked Actions tab, select the action(s) you want to remove from the list.

  2. Click the Remove Action button located next to the Link Action button.

  3. Confirm the removal if prompted.

  4. The selected action(s) will be removed from the rule.

Sort Actions by Name, Type, and Status

  1. In the Linked Actions tab, go to the Linked Actions list.

  2. Click the Action Name column header. The list sorts alphabetically in ascending order.

  3. Click the column header again to sort in descending order.

    Similarly, you can sort by Action Type and Status.

Use Case

A contact center wants Copilot to proactively assist agents when a customer shows intent to cancel a service, helping improve retention.

  1. Create a Rule: Create a rule named Retention_Risk_Detection to identify customers who may be at risk of canceling a service. Configure the rule with conditions so it is triggered when a customer message contains keywords such as “cancel my subscription” or “close my account,” or when the interaction sentiment is negative. When these conditions are met, define actions for the rule to execute, such as displaying a Copilot guidance card with retention suggestions, surfacing a relevant knowledge article like Retention Offers and Discounts, and triggering a task assist workflow that prompts the agent to offer a retention discount or similar option.

  2. Assign the Rule to a Copilot for Agents Profile: After creating the rule, assign it to a Copilot for Agents profile so it is applied during agent interactions.

    For example, link the Retention_Risk_Detection rule to the Billing Support Agents Copilot profile by navigating to the profile configuration, opening the Rules section, selecting the rule, and saving and publishing the profile. This ensures that all agents associated with this profile automatically benefit from the rule during their interactions.

  3. Rule Execution During Interaction: During a live interaction, the rule is evaluated in real time based on incoming customer messages. For instance, if a customer says, “I want to cancel my subscription. This service isn’t helpful,” the system processes the message using sentiment analysis to detect the keyword “cancel my subscription” and identify negative sentiment. The Rules engine evaluates these inputs, determines that the rule conditions are satisfied, and automatically triggers the rule without requiring any manual action.

  4. Copilot Response: Once the rule is triggered, Copilot immediately assists the agent within the interaction interface by displaying a guidance prompt indicating that the customer may be at risk of churn and suggesting appropriate retention strategies. It also surfaces the Retention Offers and Discounts knowledge article for quick reference and recommends a task assist workflow, such as applying a discount or offering an upgrade, enabling the agent to respond effectively and consistently.

  5. Outcome: As a result, the agent is able to respond more quickly with relevant retention options, improving efficiency and increasing the likelihood of retaining the customer. The rule-driven approach also ensures consistent handling of churn-risk scenarios across all agents, leading to better customer experience and standardized business outcomes.