Zia Artificial Intelligence

Zia, Zoho's artificial intelligence (AI) agent, provides comprehensive assistance to your service desk operations. Zia can be trained to learn from application data and perform various tasks to improve the service desk's performance.

You can communicate with Zia through a conversational interface in simple English.

In ServiceDesk Plus Cloud, Predictive Features and GenAI Features are essential parts of Zia AI.

• Predictive Features use application data to automate tasks such as suggesting templates, categories, subcategories, items, priorities, assigning groups and technicians to requests, and analyzing sentiments.

• GenAI Features leverage advanced generative AI tools to generate code and solutions, summarize conversations, assist with text generation and replies, automate approvals, and provide contextual answers to user queries through a virtual assistant.

Predictive Features

Predictive features are powered by Zia.

This document explains how to set up, train, and use Zia AI for prediction:

• Setup Zia AI Prediction
• Training Zia AI
• Request Prediction Features
  • Template Prediction
  • Category Prediction
  • Subcategory Prediction
  • Item Prediction
  • Priority Prediction
  • Group Prediction
  • Technician Prediction
  • Problem Prediction
  • Sentiment Analysis 
• Problem Prediction Features
  • Technician Prediction
• Change Prediction Features
  • Risk Prediction
• Solution Prediction Features
  • Solution Assist
• Limitations

Setup Zia AI Prediction 

Role Required: SDAdmin

Before enabling the prediction feature, Zia AI is shown only to SDAdmins whose application's language is set to a supported language. After enabling the prediction feature, Zia AI will be available to SDAdmins across other language setups.

Enable Prediction

• Go to Setup > Zia > Artificial Intelligence > Predictive Features.

• Select the required module.

• Use toggle to enable the required features.

• Zia offers prediction while creating, editing, or converting request type (service to incident and vice versa).

Training Zia AI

Training Zia using service desk data improves its prediction accuracy and performance.

• Training begins immediately after the predictive feature is enabled.
• Uses service desk data for initial training.
• Periodic training occurs every 7 days with the newly added or updated data.

Training Requirements

• You cannot disable predictive features while training is in progress.

• Data stored in archives and trash is excluded from training.

• If initial training fails, the system will automatically retry every subsequent day until the training is successful. To re-initiate the training manually, turn the prediction feature OFF and then ON again.

• All actions related to enabling, disabling, or training Zia AI are logged under Setup > Data Administration > System Log.

Request Prediction Features

Zia offers suggestions for requests that are created through email, web form, preventive maintenance tasks, and V3 API. With adequate training, Zia can be advanced to auto-apply certain predictions.

To access the request prediction features,

• Go to Setup > Zia > Artificial Intelligence > Predictive Features.

• Select Requests from the drop-down.

• Enable the required feature.

Template Prediction

Suggests the relevant request template based on the subject and description when a request is created, edited, or its type is converted into an incident or service.

Prerequisite

View the prediction accuracy by clicking the Prediction Rate on the Template Prediction card.

Category Prediction

Suggests the top three relevant categories based on the request's subject and description when a request is created or edited.

Prerequisite

After Zia is trained properly, click and enable Auto Apply Prediction to auto-apply the predicted category when requests are created.

Subcategory Prediction

Suggests the top three subcategories based on the subject, description, and category when a request is created or edited.

Prerequisite

Item Prediction

For requests with category and sub category, Zia provides suggestions for items.

The top three items are suggested based on the request's subject, description, category, and subcategory. Suggestions are shown when a request is created or edited.

Prerequisite

Priority Prediction

Suggests the top three priorities based on the request's subject, description, impact, and urgency when a request is created or edited.

Prerequisite

After Zia is trained properly, click and enable Auto Apply Prediction to auto-apply the predicted priority when a request is created.

Group Prediction

Zia analyzes request data and assigns the relevant technician group to the request.

During training, Zia analyzes the interdependent relationships between the technician group assigned to a request and the request details such as subject, description, and category. Zia then uses this relationship history to allocate the right technician group to incoming requests.

Prerequisite

After successful training, Zia suggests the top three groups based on the request's subject and description. Suggestions are shown when a request is created, edited, or when assigning technicians or groups from the right pane.

After Zia is trained properly, click and enable Auto Apply Prediction to auto-apply the group when a request is created.

Technician Prediction

Zia analyzes request data and assigns a technician with relevant skills to the request.

During training, Zia analyzes interdependent relationships between the technician assigned to a request and the request details such as subject, description, group, and category. Zia then uses the relationship history to allocate the right technician to incoming requests.

Prerequisite

After successful training,

• Zia suggests the top three technicians who are potentially the right fit to handle the request when requests are created or edited.

• Zia's suggestions will be listed in the technician field drop-down in the Add/Edit Request form and in the right pane.

• Suggestions are provided dynamically based on the request's subject, description, category, and group.

You can track the source of technician assignments in the request history.

While suggesting technicians, technician availability will be calculated using the Due by Time of the request, by default. If technician auto-assign is enabled, the configured technician availability model will be considered.

Auto-apply Predicted Technicians

You can enable Zia to assign the predicted technicians to the request.

Go to Setup > Automation > Technician Auto Assign and choose Artificial Intelligence (Zia) as the technician auto-assign model.

When a request is created, Zia will assign the relevant technician to the request instantly. You can find the details in the request history.

When Zia is unable to make suggestions, the Load Balancing technique will become the fallback model in allocating the technician to the request. However, if you disable technician prediction on the Zia Artificial Intelligence page, the technician auto-assign model will automatically switch to Round Robin.

Problem Prediction

Use Problem Prediction to monitor incident requests and detect emerging trends or spikes in similar incidents over a short span of time. Receive alerts on potential problems before they escalate and take necessary actions and maintain seamless operations.

Enable Problem Prediction

• Go to Setup > Zia > Artificial Intelligence > Predictive Features.

• On the Problem Prediction card, use the toggle button to enable or disable the feature.

• On enabling the feature, both the Manual and Automatic modes of problem prediction will be enabled.

• You will be directed to configure problem auto prediction.

Alternatively, after enabling the problem prediction, you can click Configure on the card to set up or edit the configuration.

Configure Problem Auto Prediction

• Incoming Request Threshold: Set a threshold for requests after which a problem identification should be triggered. Choose a threshold between 25 and 250. For instance, if you set the threshold to 200 requests, the application will start to look for potential problems when 200 incident requests are received. The application will also suggest the recommended threshold based on request inflow in the last two months.

• Technicians to Notify: Select the technicians to be notified when a potential problem is identified. Only technicians with edit permissions for problems and requests will be listed. You can select a maximum of 100 technicians.

• Click Save.

Run Problem Predictions Manually

After enabling the problem prediction, the following options are displayed in the Quick Actions header menu: Run Prediction and View Predictions.  

Run Prediction: Use this option to initiate problem prediction manually. Once you initiate this, the system will analyze the last 500 requests for any patterns and trends and send alerts on potential problems. To start the prediction, click Run Prediction and choose the required request filter. For instance, you can choose the My Pending Requests filter to predict problems from requests and click Run.

After the prediction is completed, you will receive a bell notification. Clicking it will take you to the predicted problems list view.

If no potential problems are detected, the following screen will be shown.

View Prediction: Lists all the problem predictions performed by the system. Click a prediction to view its details.

Requests predicted with potential problems can be associated with existing problems or with a new problem.

Sentiment Analysis

The Zia Sentiment Analysis feature examines request conversations to determine whether the emotion expressed is positive, negative, or neutral.

Zia analyzes the first 2000 characters of a requester's conversations and places appropriate emojis. Then, the overall sentiment score is calculated and displayed in the right panel of the request details page.

Uses of Sentiment Analysis

This feature enables technicians to:

• Prioritize requests.
• Send personalized responses.
• Assess user satisfaction.

Enable Zia Sentiment Analysis

• Go to Setup > Zia > Artificial Intelligence > Predictive Features.
• Enable Sentiment Prediction.

View Analysis Details: Shows the number of conversations and sentiments predicted by Zia, along with the overall sentiment score. The score will be displayed even if the sentiment prediction is disabled.

The sentiment score is calculated using sentiment points.

Sentiment Points

Positive - 1

Negative - 0

Neutral - 0.5

Formula to calculate sentiment score

Overall_sentiment_score = [(positive_sentiment_count + (neutral_sentiment_count*0.5)]/total_sentiment_count) * 100 

Sentiment Score

0%-30% - Dissatisfied - 

31%-60% - Neutral -

61%-100% - Satisfied -

You can view the overall sentiment score and the sentiment of the recent conversation from the right pane of the request details page. Hover over the overall score to view the emotion of each conversation.

Problem Prediction Features

To access the problem module prediction features,

• Go to Setup > Zia > Artificial Intelligence > Predictive Features.

• Select Problems from the drop-down.

• Enable the required feature.

Technician Prediction

Analyzes data in the Problems module and assigns technicians with relevant skills to problems.

During training, Zia studies the interdependent relationships between the technician assigned to a problem and details such as subject, description, group, and category, with the first three fields taking equal precedence over category. Zia then uses this relationship history to allocate the right technicians to new problems.

Prerequisite

If any sites do not contain the necessary number of problem requests with assigned technicians, the training will fail for that particular site, and the prediction rate for the failed site will be null. Zia will then attempt to train the failed sites in subsequent trainings. Click the Prediction Rate on the technician prediction card to view the prediction rate of each site.

Zia suggests the top three technicians based on the subject, description, group, and category when a problem request is created or edited. Zia's suggestions are listed in the Technician field drop-down in the Add/Edit Problem form and in the right panel.

Change Prediction Features

To access the change prediction features,

• Go to Setup > Zia > Artificial Intelligence > Predictive Features.

• Select Changes from the drop-down.

• Enable the required feature.

Risk Prediction

Zia suggests the top two risks of change requests based on their title, description, priority, impact, urgency, change type, and emergency when a change request is created or edited.

Prerequisite

Solution Prediction Features

Solution Assist

Zia summarizes content from one or more solution articles and provides solutions to user queries in the Zia chatbot.

• Go to Setup > Zia > Artificial Intelligence > Predictive Features.

• Select Solutions from the drop-down.

• Enable Solution Assist.

When users type in a query in the chatbot, Zia automatically parses and summarizes the relevant solution content as a response in the chatbot. Click Explore Related Solutions  to view the related solution articles.

Include the Zia solution suggestion variable ($ZiaSolutionSuggest) in the request notification template to send the predicted solution for the reported issue as an initial response to the requester.

Limitations

Currently, Zia AI has the following limitations that will be improved over time.

Language Limitations

• Zia AI can be trained to analyze request data and predict request fields, templates, and technicians only in the supported language setups.

• Problem prediction, sentiment analysis, solution assist, and auto open/close of requests are supported only in English setup.

• If the last SDAdmin changes their application language to an unsupported language, the Zia AI configuration page will be disabled automatically for all SDAdmins.

Data Deletion Limitations

• Data used by Zia for training will be deleted from the machine-learning server only after Zia is disabled. If a particular feature is disabled, the relevant data will also be deleted from the machine-learning server automatically.

• If an evaluator enables and trains Zia, the training data will be deleted when the organization created by the evaluator is deleted.

Other Limitations

• Zia AI prediction will not work on requests created via V1 API.

• See Also: How to open and close requests automatically using Zia Artificial Intelligence.