Skip to main content
The dashboard in the AI control center provides a compact overview of the performance of the deployed AI agents.It combines basic quality indicators with detailed performance metrics, allowing for a quick assessment of the reliability and efficiency of the current automations.

​
Basic AI Metrics

These key figures provide information on how reliably the AI operates in daily use.
  • Customer correctly identified
    The proportion of work sessions in which the customer could be assigned unequivocally and without manual intervention.
  • Correctly categorized with tags
    The proportion of work sessions in which the AI set the correct tags. No manual intervention was required.
  • Text Assistant Accuracy
    Average accuracy of the text assistant, calculated across all work sessions in the selected period.

​
AI Performance Metrics

This is where the operational effectiveness of the automation becomes apparent:
  • Automatic Processing
    Indicates the number of tickets that were processed automatically, distinguished by With Approval (L4) and fully autonomous (L5).
  • Degree of Automation
    Evaluates how strongly tickets were overall supported by the AI:
    • L0: No AI, except identification/categorization
    • L1: AI used for text editing
    • L2: Adopted the text suggestion of the basic agent
    • L3: Suggestion of a specialist agent confirmed
    • L4: Dark processing with approval
    • L5: Complete dark processing

​
Agent Overview

The table view lists the AI agents with active automation, i.e., those that process tickets with approval (L4) or fully autonomously (L5). For each of these AI agents, the following are displayed:
  • Processed: Number of tickets processed during the period
  • Awaiting approval: Cases currently awaiting manual approval
  • Automatically processed cases (L4+L5): Total of all successfully automatically closed tickets
  • Dark processing rate (L5): Proportion of cases that were closed without human intervention
  • Human approval rate (L4): Proportion of cases where an employee approved the AI proposal
  • Error rate: Proportion of cases that had to be corrected or withdrawn