Detect sensitive topics by using sensitivity detection in Now Assist for HR Service Delivery (HRSD)

  • Release version: Xanadu
  • Updated October 30, 2024
  • 1 minute to read

    • Detect and fix sensitive topics that relate to your employee relation cases by using the sensitivity detection capability in the Now Assist for HR Service Delivery (HRSD) application.

    Before you begin

    Role required:
    • sn_hr_gen_ai.admin
      • sn_hr_core.case_writer
      • sn_hr_core.case_reader
      • sn_hr_er.case_reader
      • sn_hr_er.case_writer

    About this task

    This capability helps you detect sensitive cases such as the harassment complaints, discrimination allegations, workplace violence, safety, employee behavior, and employee personal issues. You can also block Now LLM Service from engaging with these sensitive case types.

    Note:

    The store or app client should automatically take care of all the dependencies when you install Now Assist for HRSD. You must install Now Assist for HRSD version 5.0.7.

    Procedure

    1. Navigate to the table sys_gen_ai_filter and set active to true for the filters.
    Note:
    sn_hr_gen_ai.admin is required to perform the first step of setting active to true for the filters.
    Figure 1. Generative AI Filters
    Allows you to set active to 'true' for AI filters
    1. Verify that the following system properties are set to true.
      • com.glide.genai.semantic_filter.enabled
      • com.glide.cs.semantic_search.enabled
    2. Navigate to the sys_one_extend_capability table and select the VA skill search capability.
      In the capability attributes, set the apply filter field for the utterance to be true.
      Note:
      Repeat this step for the user_message attribute in the Virtual Agent free-form capability and the search_text attribute in Virtual Agent Semantic Search.
    3. Navigate to Now Assist Admin Console > Features > Platform > Conversation Interfaces.
    4. Set up Now LLM Service.
    5. Navigate to Virtual Agent on one of the portals and enter Sensitive Phrases.
      Note:
      You enter Sensitive Phrases at the beginning of a conversation. This action should trigger the Sensitivity Detection Fallback topic. When sensitive phrases are entered in the now assist panel, it should be detected as sensitive and the LLM should not respond and should allow you to redirect to a live agent or create a case.

    Result

    The Sensitive Phrases are automatically detected so that your group can properly handle sensitive cases.