View Task Intelligence Analytics
Summarize
Summarized using AI
This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.
Summary of View Task Intelligence Analytics
The Task Intelligence Analytics dashboard in ServiceNow Customer Service Management (CSM) enables you to monitor and evaluate the performance of AI-driven categorization and prediction models. It provides key insights into how models assist agents by tracking prediction acceptance, replacement, and impact on case resolution times.
Show less
Key Features
- Number of Predictions: Visual line graphs display the count of correct predictions accepted by agents over time across all active models, including sentiment and language models if enabled.
- Mean Time to Resolve (MTTR): Tracks the average time to resolve cases, helping you assess how model predictions contribute to faster case handling.
- Prediction Utilization Monitoring: Detailed widgets show accepted, replaced, and skipped predictions by agents for specific models and output fields, indicating model effectiveness and areas for retraining.
- Performance Overview Table: Presents mean percentage values of correct, incorrect, and skipped predictions for each model-output field combination, facilitating quick performance assessment.
- Prediction Usage Tracking: Bar charts track accepted, replaced, and skipped predictions over time, with options to view counts or percentages and compare against training data baselines.
- Multi-Model Comparison: Supports multi-select for model and output fields to compare agent usage and training records across different models and prediction types.
- Persistent Configuration: Dashboard selections, including models, fields, usage types, and display options, are retained on page refresh for consistent analysis sessions.
Practical Application
Access the Task Intelligence Analytics page via All > Task Intelligence for Customer Service > Monitoring. You can analyze either all models collectively or select individual models created through the Task Intelligence Admin Console. This allows you to:
- Monitor how well your AI models are assisting agents by measuring accepted versus replaced predictions.
- Identify declining usage trends to determine when a model retraining is necessary.
- Evaluate the business impact through MTTR improvements linked to prediction assistance.
- Customize views to focus on specific models, output fields, or time periods.
- Compare performance across multiple models to optimize AI-driven case categorization and resolution.
This analytics capability empowers ServiceNow customers to continuously improve AI model effectiveness, enhance agent productivity, and drive better customer service outcomes by making data-driven decisions about model training and deployment.
View the Task Intelligence for CSM Analytics dashboard to monitor the model performance over time, track the business value, and see what predictions your agents did or didn't use.
The dashboard uses visual representations to provide you with an overview of how each model is performing. You can see the number of predictions from each model and the overall mean time that it takes to resolve the cases in your organization.
The following example shows the Get an overview and See how your trained model is doing sections within the dashboard.
Get an overview
The dashboard contains visual representations that provide you with the business value of the model.- Number of predictions
- The line graph shows the number of correct predictions that agents have accepted over time for all the categorization models combined. If the sentiment and language models are activated, the correct predictions for those models are shown as well. As the model continues to learn, it can increase the number of predictions that agents accept.
- Mean time to resolve (MTTR)
- The line graph shows the average amount of time that it takes to resolve cases over time. As the model makes more predictions, the MTTR should decrease as the predictions help your agents.
The Task Intelligence Analytics page includes the following data about the predicted fields and how they impact your business:
- The number of predicted fields.
- The number of predictions that agents used.
- The number of predictions that agents changed.
See how your trained model is doing
The dashboard uses visual representations to help you monitor the utilization of predictions by the model over time. By selecting a specific model and output field, you can view the number of predictions that were accepted or
replaced by the agent and compare them to the baseline.
- Predictions agents accepted
- The widget shows the correct predictions that your agents used during case management over time. If this number is trending downward, you can look to retrain your model.
- Predictions agents replaced
- The widget shows the incorrect predictions that your agents removed during case management over time. If this number is trending upward, you can look to retrain your model.
- Predictions the model skipped
- The widget shows the number of predictions that were skipped by the model based on the model, output field, and date range selection.
- Performance overview
- The performance overview table shows the mean percentage values for correct, incorrect, and skipped data for each combination of model and output field.
- Track usage of individual field predictions over time
- The bar chart tracks the usage of the individual field predictions over time. Each bar in the chart shows three components, which are the predictions accepted, the predictions replaced, and the predictions that were skipped by the model. A red outline around each bar represents the total number of records for each day. To compare specific components, navigate to the legends and deselect the ones that you don't want to include so that you can have a more customized and focused comparison based on user preferences. By default, the view displays the number of predictions. However, you have the option to switch to the percentage view by toggling the Show Percentage option. In the percentage view, you can also access the information about the baseline that was replaced and accepted, which is derived from the training data. This option enables you to gain insights into the performance of the model with the baseline.
- Track usage of model predictions over time
- The bar chart monitors the usage of the model over time and is similar to the bar chart in the “Track usage of individual field predictions over time” data. The field selections enable users to compare predictions over time across a maximum of three fields.
Note:
- When you refresh the page, the configurations, model selection, fields, usage type, and Show percentage toggle selection are retained.
- The Model and Output field options support multi-select functionality so that you can compare the average number of training records and analyze how your agents use the predictions of different models. When multiple models are selected, the Compare how agents used models' predictions section enables your agents to compare the selected models across different usage types. The chart also supports toggling between the count and percentage. It also displays the baselines for accepted and replaced predictions when you’re using the percentage model.
Using Task Intelligence analytics
Navigate to to access the Task Intelligence Analytics page.
You can view the analytics for all of your models or each individual model. To select a model, do the following actions:
- Select Model to access analytics for an individual model.
- Select a model from the list of models that have been created from the Task Intelligence Admin Console.
- Select Apply.