Zia Configurations  

Zia configurations include various AI-driven actions for request analysis, prediction, and user interaction, such as:

• Validate request reopening
• Assign/suggest category prediction
• Suggest template prediction
• Zia parser
• Execute user approval action

These actions are enabled by default under Admin > Zia Configurations > Configuration. To disable them, turn off the toggle next to required action.

You can train these actions by adding your business real-time data and test the data to make sure that the results align with your operational objectives.

Technicians can provide feedback to Zia's predictions. Continuous feedback improves accuracy and enables the features to adapt to your workflows. 

For more details, refer to the following sections:

• Customize training data
• User feedback to Zia's predictions
• Track Zia's accuracy
• Track history

Validate Request Reopening 

Zia evaluates end user responses to completed requests and reopens only requests that require further action. This rule applies to requests in any completed status, including default and custom statuses.

This prevents the help desk from being swamped by automated responses, such as thank-you notes, acknowledgment messages, out-of-office emails, and the like. Consequently, technicians can focus on resolving more critical requests within the defined service level agreements.

Use the Customize Data option to improve Zia’s accuracy by adding custom training data based on your operational objectives. Learn more.

Suggest  Category/Template Prediction 

Zia analyzes a request's subject and description and suggests the appropriate category and template during request creation. This helps users select the correct category and template, reducing the technicians' effort spent in assigning them manually. Users can accept or reject Zia's suggestions as needed.

Suggestion shown in the request web form

• If the requester rejects a suggestion, Zia will recommend the suggestion on the request details page to other users who have edit permissions and did not reject the recommendation earlier. These users can accept or reject Zia's suggestion, or recommend the template to the requester. 
• Once the request owner (technician assigned to the request) rejects the recommendation in the request details page, it will no longer be shown to other users for that request. 

Training Zia for Category and Template Prediction  

• Zia requires a minimum of 500 and a maximum of 50,000 requests to be trained for the prediction. Every night, Zia retrains with the latest 10,000 records.
  • When Zia starts training, SDAdmins/technicians will be notified via push/bell notification. 
  • If there are more than 50,000 requests, Zia uses the latest 50,000 requests for training.  

Zia Parser 

Zia parser analyzes incoming request emails to perform various request actions, such as creating requests, updating request properties, and more. However, actions remain subject to the email sender's permissions.

For Zia to perform parsing, incoming emails must contain an identifier in their subjects.

You can configure Zia parser as follows:

• Specify the parsing identifier (Default Identifier: @ZIA). 
• Choose which fields or actions the parser should process. 
• Customize notifications related to Zia parser and parsing results.
• Use Test Zia Parser to verify the setup.

Learn more.

Execute User's Approval Action 

Zia analyzes the approver's response to approval emails, deduces approval decision, and applies the decision on requests, changes, releases, purchase orders, and purchase requests.

Use the Customize Data option to improve Zia's accuracy by adding custom training data based on your operational objectives. Learn more. 

You can also manage the custom training data and other approval configurations, such as email-based approvals and acknowledgments, notifications for prediction results, under Admin > General Settings > Advanced Portal Settings > Approval. 

Customize and Test Training Data    

If Zia's predictions do not match your expectations, you can enrich the training dataset by adding real-time data or by customizing existing data. Zia retrains itself with the new data every 15 minutes and expands its learning.

You can add custom training data for:

• Request Reopening Prediction: Add keywords or contextual statements that indicate whether Zia must reopen requests or retain their closure. 
• Approval Prediction: Add phrases that help Zia predict whether an approval must be approved or rejected, and recognize whether request approvals require further clarification.  

To do this,

• Click Customize Data under the respective Zia prediction options. 

In the training dataset pop-up,

• Use the editor of the required action panel to add, edit, or delete data.
• Click Add New Data at the top of the required panel to enter a new statement at the end of the list. 

• Use the Test Zia to Predict Results option to verify whether Zia predicts relevant actions for sample phrases. 

To remove all your custom data and revert to the default training dataset, click Restore Default at the top of the pop-up. 

Dataset Settings (Approval Predictions Only)

Training data added in one language works across other languages with up to 80% accuracy. If prediction is inaccurate for a particular language, you can improve the accuracy by adding training data in that language. To do that,

• Select For Specific Language under Show Dataset. 
• Choose the required language. 
• Add sample statements in the required dataset panels. 

You can also derive training data from user replies to approval emails by selecting Added by email under Show Dataset. You cannot modify this data or add new ones, but can delete them. 

• Finally, click Submit or Submit and Close to save the changes.

User Feedback to Zia Predictions 

Zia's prediction results of request reopening and approvals are notified in the Zia chat bot on the footer and Zia Notifications on the <module> details page. Zia parser results are notified in Zia Notifications on the corresponding request details pages.

Technicians can review these results and share their feedback (upvote /downvote ). Zia uses this feedback to re-equip its training data regularly and improve accuracy rate.

Role Required to Share Feedback:

• Technicians with the ZiaVerifier role can provide feedback to Zia's predictions. 
• By default, SDAdmins are assigned the ZiaVerifier role. 
• Other technicians can be assigned the ZiaVerifier role for a module if they have view permissions to that module. To assign it, use the Assigned Roles drop-down on the technician details page. 

Illustrations of Zia Notifications on the request details page: 

Approval Prediction

Reopen Prediction 

Parser Results

You can manually train Zia for inconclusive and incorrect predictions on request reopening and approval actions. To do that,

• Go to Zia Notifications or Zia chat bot.
• Go to Reopen, or Approval tab. 
• Perform the following: 

Track Zia's Accuracy    

Zia's accuracy rate helps you track its learning over time. The accuracy rate is computed daily at 11:55 PM and displayed next to each Zia prediction action.

Track History    

In the Zia Configurations page, click View History on the top right to view a log of operations performed on Zia configurations.