Predictive Intelligence: How can I block a certain HR service from being updated

[tworzala]

I have predictive intelligence on to classify the HR service correctly. It keeps incorrectly classifying one of our services and I would like to block PI from changing certain services. Is there a way to do that?

[Itallo Brandão]

Hi tworzala,

This is a common scenario. Predictive Intelligence is triggered by a Business Rule (or sometimes a Flow) that runs on the Case table. To "block" PI from updating specific scenarios, the best practice is to modify the Trigger Condition of that Business Rule so it skips those records.

Here is how to do it:
1. Find the Trigger Business Rule

• Navigate to System Definition > Business Rules.

• Search on the Table where your cases are being created (e.g., sn_hr_core_case or sn_hr_core_case_creation).

• Look for a rule with a name related to "Prediction", "Classification", or "Predict HR Service". (Common OOB names are like Default HR Service or Predict Field Value).

2. Add an Exclusion Condition

• Open the Business Rule.

• In the When to run tab (or the Filter Conditions section), add a condition to exclude the specific service you want to protect.

• Condition: [HR Service] [is not] [The Service You Want to Block]

3. Best Practice: Prevent Overwriting

• Usually, we only want PI to predict if the field is empty. Ensure your Business Rule condition also includes:

• AND [HR Service] [is empty]

• If this condition is present, simply manually setting the HR Service during creation (or via Record Producer) will automatically stop PI from overwriting it.
  
  

By adding these conditions, the system will verify the record before calling the Predictive Intelligence API. If it matches your exclusion criteria, the prediction is skipped entirely for that case.

Hope this helps you regain control over that service!

If this response helps you achieve your requirement, please mark it as Accepted Solution.
This helps the community grow and assists others in finding valid answers faster.

Best regards,
Brandão.

[tworzala]

I did that but that doesn't seem to do anything. I also excluded those cases from the training but that hasn't seemed to fix the issue either. Essentially, I want to say "don't let PI predict x,y, z service". Is that possible?

[Itallo Brandão]

Hi @tworzala ,

Thank you for the clarification.
If your goal is to explicitly "Blacklist" specific services (X, Y, Z) so they are never predicted as an output, simply excluding them from the Business Rule trigger won't work (that only controls when it runs, not what it predicts).

To completely stop PI from predicting "Service X", you must ensure "Service X" does not exist in the training model. If the model hasn't learned "Service X", it mathematically cannot predict it.

Here is the correct process to "Blacklist" those outputs:

1. Update the Solution Definition Filter

1. Navigate to Predictive Intelligence > Classification > Solution Definitions.

2. Open the specific definition used for your HR Services (e.g., ml_sn_hr_core_global_hr_service).

3. In the Filter section, add logic to explicitly exclude the services you don't want:

   • [HR Service] [is not] [Service X]

   • AND

   • [HR Service] [is not] [Service Y]

4. Why: This tells the system: "When learning patterns, ignore any historic examples where the result was Service X."

2. The Critical Step: Update & Retrain Changes to the filter do not apply instantly. The current active solution in your instance still has the "Old Brain" that remembers Service X.

• You MUST click the Update & Retrain button.

• Wait for the training to complete.

• Once the new solution version is Active, the system will effectively have "forgotten" that Service X exists and will stop predicting it entirely.

3. Troubleshooting "I already excluded them" If you believe you already did this step, check the following:

• Did you Retrain? If you just saved the filter but didn't run a full training cycle, nothing changed.

• Check "Class Confidence": Alternatively, if you don't want to retrain, you can go to the specific Class Confidence record for that solution and set the threshold for "Service X" to 100 (impossible to reach), effectively disabling it. But the Retrain method above is cleaner.
  
  

If this response helps you solve the issue, please mark it as Accepted Solution.
This helps the community grow and assists others in finding valid answers faster.

Best regards,
Brandão.

[tworzala]

I had already done those step. I excluded the HR services I do not want it to predict and retrained. In the new Class Confidence, that HR service does not appear and it doesn't appear in the Excluded Classes. This tells me that there's something else running. I already inactivated the other solution definitions. Any other ideas? In the work notes it does say "Auto-categorized case with predicted COE: sn_hr_core_case_workforce_admin, predicted Service: Unemployment Claim Data Request" which tells me this is PI.