Using Predictive Intelligence
Summarize
Summary of Using Predictive Intelligence
Predictive Intelligence in ServiceNow enables you to train AI models on the ServiceNow AI Platform that predict, estimate, and identify patterns within your data. These models support various practical tasks such as routing work, populating form fields, estimating wait times, and more. Key use cases include assigning, categorizing, and prioritizing tasks, detecting major incidents, recommending case resolutions, preventing duplicates, and detecting phishing attempts. The solution integrates seamlessly with other ServiceNow products like Document Intelligence and Task Intelligence.
Show less
Training and Extending Models
You can create and train machine learning solutions using your instance data to predict, recommend, and organize information. Predictive Intelligence supports extending capabilities to specific processes like incident categorization and customer service management (CSM) case assignment, enhancing automation and accuracy in these areas. Training solutions involves defining the problem, preparing quality historical data, and using ServiceNow frameworks without needing to write code.
Testing and Monitoring
After training, you invoke the Predictive Intelligence API to generate predictions and evaluate model performance. The Solution Statistics dashboard provides essential metrics such as:
- Average and Daily Prediction Coverage: Measures the percentage of records where predictions were made.
- Average and Daily Prediction Precision: Measures the accuracy of predictions compared to actual outcomes.
These insights help you monitor model effectiveness and guide ongoing improvements.
Preparing Your Instance
For optimal results, prepare by identifying specific problems to solve with Predictive Intelligence and ensuring you have 30,000 to 300,000 high-quality records for training. Consistent and complete data is crucial to avoid unreliable predictions.
Implementation Process
The typical implementation spans approximately 14 days, involving:
- Day 1: Clone production to a non-production instance.
- Days 2–10: Create, train, and validate solution definitions on the non-production instance.
- Days 11–13: Transfer solutions to production, retrain, validate, and configure retraining schedules.
- Day 14 and beyond: Monitor and maintain the deployed solutions.
Using non-production environments allows you to safely test workflows and fine-tune models before full production deployment.
Train and use Predictive Intelligence solutions to accomplish various tasks and that integrate with other ServiceNow products, such as Document Intelligence and Task Intelligence.
Overview of Predictive Intelligence
Predictive Intelligence is the interface by which you can train models on the ServiceNow AI Platform. These models enable you to predict, estimate, and identify patterns that can be used to route work, populate form fields, estimate wait times, and more.
- Show suggestions for relevant articles.
- Assign, categorize, and prioritize tasks.
- Detect major incidents.
- Recommend case resolutions.
- Prevent duplicate articles and ideas.
- Detect phishing attempts.
For more information about the different types of solutions available, see Explore Predictive Intelligence.
Training your ML solutions
Predictive Intelligence enables you to train predictive models and machine-learning solutions that you can apply using data on your instances. The solutions you create use the frameworks to predict, recommend, and organize data. To get started, see Creating and training solutions.
- Incident categorization: Predicts the incident category based on the short description. See Predictive Intelligence for Incident Management.
- CSM case assignment: Predicts the case record assignment group based on the short description. See Predictive Intelligence for case management.
For more information, see ServiceNow apps and features that use Predictive Intelligence.
Testing and monitoring predictions
After creating and training your solutions, call on the Predictive Intelligence API to make a solution prediction. Use the results to gauge the performance of the solution and make changes as needed.
| Report | Description |
|---|---|
| Average Prediction Coverage (last 30 days) | The percentage of predictions that yielded an outcome out of the total number of predictions attempted. Click the coverage score to see a breakdown by class. |
| Daily Prediction Coverage | The percentage of records created on a given day in which the solution was able to predict an outcome. |
| Average Prediction Precision (last 30 days) | The percentage of predictions in which the predicted value was the same as the final value of the field when the record closed. Click the precision score to see a breakdown by class. |
| Daily Prediction Precision | The percentage of records closed on a given day in which the predicted field value was the same as the final value. |
For more information, see Testing and monitoring predictions.
Preparing your instance
For you to get the most out of Predictive Intelligence, you will want to prepare. You don't need to write code or do calculations, but deciding what you hope to do with the solution definitions will make implementation easier.
- Identify the problems that you want to solve with Predictive Intelligence.
- Have 30,000–300,000 high-quality records from which Predictive Intelligence can learn.
- Set your expectations.
Implementation Process
Predictive Intelligence takes approximately 14 days to implement on a production instance.
- Day 1: Clone production instance over to a non-production instance.
- Days 2–10: Create a solution definition, train it on historical records, and validate that the solution works as desired on the non-production instance.
- Days 11–13: Create import and update sets to move the solution to production, train and validate on the new instance, and set the retraining frequency.
- Day 14 and on: Monitor the solution.
In general, non-production environments are where workflows can be tested and formatted before being moved over to the production instance to further train models and test predictions.
For more information about getting started with Predictive Intelligence, see our guide on how to get started with Predictive Intelligence.