- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-19-2023 12:38 AM
Hello Everyone,
I have a requirement to auto-populate the assignment group in the incident table using PI (Predictive Intelligence).
What are the best practices for implementing Predictive Intelligence (PI)?
I have created a classification solution definition with the data and trained the system in a non-production environment. However, I am concerned about how it will work in the production environment.
1. Do I need to develop and train PI directly in the production instance, or can I do it in a non-production instance and then move it to the production instance?
2. If I am doing this in a non-production instance, how can I move my developments to the production instance, especially the machine learning training data?
Business Rules – Used in Predictive Intelligence.
What is the function of the “Update Prediction Results” business rule? Does it only update the Predicted Correctly, Final Input Value, and Final Output Value fields in the [ml_predictor_results] table, or does it also store updated records (values) in other tables?
Solved! Go to Solution.

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-19-2023 03:56 AM
Hi Sathya, please see my responses below.
1. Do I need to develop and train PI directly in the production instance, or can I do it in a non-production instance and then move it to the production instance?
(Lener) You can do your training on non-production. Then move it to production.
2. If I am doing this in a non-production instance, how can I move my developments to the production instance, especially the machine learning training data?
(Lener) You can either use an update set to move the PI solution or you can export the XML. Regarding moving training data. The recommendation is to re-train your solution definition on production.
3. What is the function of the “Update Prediction Results”
(Lener) Correct it's core function is to populate the ml_predictor_results table which ties to a PA dashboard that you can use to monitor the effectiveness of your classification solution.
You may also wish to install Task Intelligence for ITSM. This is a new AI interface for ITSM classification meant to simplify the creation, testing, and deployment of AI models on ServiceNow. Also check out the AI Community > Advanced Topics for best practices and tuning tips. -Lener

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-19-2023 03:56 AM
Hi Sathya, please see my responses below.
1. Do I need to develop and train PI directly in the production instance, or can I do it in a non-production instance and then move it to the production instance?
(Lener) You can do your training on non-production. Then move it to production.
2. If I am doing this in a non-production instance, how can I move my developments to the production instance, especially the machine learning training data?
(Lener) You can either use an update set to move the PI solution or you can export the XML. Regarding moving training data. The recommendation is to re-train your solution definition on production.
3. What is the function of the “Update Prediction Results”
(Lener) Correct it's core function is to populate the ml_predictor_results table which ties to a PA dashboard that you can use to monitor the effectiveness of your classification solution.
You may also wish to install Task Intelligence for ITSM. This is a new AI interface for ITSM classification meant to simplify the creation, testing, and deployment of AI models on ServiceNow. Also check out the AI Community > Advanced Topics for best practices and tuning tips. -Lener
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
10-19-2023 04:39 AM
Thank you for your response @Lener Pacania1 .