Recommended Agents: Five Main Pillars

This document provides a structured overview of the five key pillars used to determine recommended agents in SupportLogic. Each pillar contributes to the overall score that determines agent suitability for case assignment.

1. Time Overlap

Time Overlap measures the alignment between an agent’s active hours and the predicted customer hours based on the reporter’s activity.

Agent Active Hours Calculation

• Primary Calculation: Assigned shift hours minus Out-of-Office (OOO) time.

• Fallback Calculation: If no working hours are found, assignment hours minus OOO time are used.

• Final Fallback: If neither working nor assignment hours exist, predictive hours minus OOO time are used.

• If an agent has multiple shifts, all assigned shifts are considered using the same logic.

Overlap Calculation

• Covers the next 24 hours from the moment of checking the time overlap.

• Weekend hours are skipped if they fall within this period.

UI Color Codes

• Dark grey: Agent and customer hours overlap.

• Red: Agent’s OOO time in the next 24 hours.

• Blue: Overlap hours between agent and customer.

• The higher the overlap, the higher the total Time Overlap percentage.

2. Customer

Evaluates an agent’s history with an account based on past case handling performance.

Factors Considered

• Average case sentiment.

• Change in sentiment while the agent was assigned.

• Need attention score.

• Resolution time (Time to close used as a proxy).

• Recency of the case.

• Final case resolver.

Scoring

• Baseline Score: 0.2 (20%) for agents with no history with the account.

• Score drops below 20% for poor past interactions.

• Score rises above 20% for positive past interactions.

UI Display

• Total customer experience percentage.

• Number of cases the agent worked on for the same account.

• Average sentiment on those cases.

• "Customer" refers to an account, not an individual reporter.

3. Skills

Determines agent suitability based on expertise and case-specific requirements.

Data Sources

• Ontology file: Extracted and maintained by SupportLogic Solution Architects (SAs).

• Case Skills/Keywords: Extracted on demand from case comments (once per customer, upon ICA module onboarding).

4. Bandwidth

Evaluates an agent’s capacity to handle new cases based on current workload.

Factors Considered

• Case Priority: Higher priority cases receive more weight.

  • Example: A case with priority 8 is treated as 8 separate cases by the model.

• Case Status:

  • New cases require more effort and receive a higher weight.

  • Certain statuses (e.g., "hold", "closed") are auto-detected and set to zero.

• Recommended Weight Range: 0 - 1.

UI Breakdown

• Displays active case backlog by status/priority.

• Shows cases awaiting assignment with priority and active escalations (if any).

Incoming Case Workload Prediction

• Extra penalty applied to agents without sufficient bandwidth.

• Weights are set at model training.

• To modify case priority weights, a Jira model request is required.

5. Requirements

Trained Bandwidth Model & Assignment Penalty

Penalty for Recent Assignment

• Prevents overload by applying a time-based penalty to agents assigned a case within the last X minutes.

• Default settings: 180 minutes with a 0.6 weight.

• The penalty weight decays over time.

  • Example: An agent assigned a case 1 min ago gets a larger penalty than one assigned 170 mins ago.

• Important: This penalty does not prevent case assignment. If the affected agent has the highest score, they can still receive the case.

Default Weights of Each Pillar

The default weights assigned to each pillar are as follows:

• Time Overlap: 0.33

• Customer: 0.17

• Skills: 0.17

• Bandwidth: 0.33
  
  

  Recommended agent cards:
  
  

  
  

  
  
  Recommended agents panel: