- Home
- /
- Article
Thanks for your feedback.
In this article
Feedback?Use this section to prepare your organization to evaluate, activate, and monitor personalized AI routing.
Before you begin
Before beginning the setup, make sure you understand how Personalized AI Routing works, have configured AI Routing, and have confirmed that your organization and queues contain sufficient historical data. You should also know which business outcome you want to improve.
| 1 |
Confirm that AI routing is enabled for the organization and that your administrator role has permission to configure routing. |
| 2 |
Identify the queues that have enough completed interactions and a measurable business outcome. |
| 3 |
Confirm that the queue uses the correct teams, skills, channel capacity, and flow logic. |
| 4 |
Choose the business outcome that you want AI routing to optimize. |
| 5 |
Confirm that outcome data is consistently captured for completed interactions. |
| 6 |
Confirm that users who will review results have access to the AI routing reports in Analyzer. |
Analyze queues for AI routing
The optimization check analyzes the selected queues for the selected business outcome. Use the result as guidance when you decide whether to evaluate or activate AI routing for a queue.
| 1 |
Go to the AI routing configuration page in Control Hub. |
| 2 |
Select the business outcome that you want to evaluate. |
| 3 |
Select one or more queues to analyze. |
| 4 |
Run the optimization check. |
| 5 |
Review the optimization status for each queue. |
| Status | What it means |
|---|---|
| High potential | The queue has enough data and expected improvement potential. Start evaluation mode before activating AI routing. |
| Low potential | The queue may have limited expected improvement from AI routing for the selected business outcome. This result is advisory and doesn't prevent you from using evaluation mode or AI routing. |
| Pending or in progress | The optimization check hasn't completed. Wait for the check to finish before you start evaluation mode. |
| Failed | The check couldn't complete. Review the selected queue, outcome data, and permissions, then run the check again. |
Evaluate AI routing
In evaluation mode, regular routing continues to handle interactions. AI routing predicts the recommended agent and expected outcome, but it doesn't change live routing behavior. Use the evaluation report to decide whether to activate AI routing for the queue.
| 1 |
Open a queue that has an optimization result. |
| 2 |
Start evaluation mode for the selected business outcome. |
| 3 |
Allow evaluation mode to run long enough to collect representative traffic for the queue. |
| 4 |
Open the AI routing evaluation report in Analyzer. |
| 5 |
Compare the predicted AI routing results with the regular routing results. |
| 6 |
Activate AI routing only when the evaluation report supports the expected outcome improvement. |
Activate AI routing
When AI routing is activated, Webex Contact Center uses AI recommendations for live interactions in the selected queue. Existing queue, skill, priority, flow, availability, and capacity rules still determine agent eligibility.
| 1 |
Review the optimization check result and confirm that the queue has high potential. |
| 2 |
Review the evaluation report for the selected business outcome. |
| 3 |
Confirm that the organization has the required entitlement to use AI routing for live traffic. |
| 4 |
Activate AI routing for the queue. |
| 5 |
Monitor the queue in Analyzer after activation. |
AI routing starts handling eligible interactions for the configured queue and outcome. If major queue configuration, staffing, skill, flow, or outcome data changes occur, run the optimization check and evaluation again.
Monitor AI routing
Monitor AI routing regularly in Analyzer. Review enough traffic before drawing conclusions, especially for queues with lower contact volume or seasonal patterns.
Metrics to review
- Interaction volume included in the evaluation or AI routing report.
- Selected KPI average for AI routing and regular routing.
- Predicted and actual outcome values.
- Percentage improvement and absolute improvement.
- Daily or weekly trend for the selected business outcome.
Reporting threshold
An interaction must be longer than 30 seconds to appear in AI routing reports. This applies to evaluation mode and AI routing mode.
When results don't improve
If AI routing doesn't improve the selected outcome, review the selected KPI, historical data quality, queue configuration, staffing distribution, and recent operational changes. Rerun the optimization check after significant changes.
Best practices
- Start with one clear business outcome for each queue.
- Use queues with enough completed interactions and meaningful variation in agent outcomes.
- Run evaluation mode before activating AI routing for live traffic.
- Keep disposition, global variable, and KPI data consistent across interactions.
- Compare AI routing reports with regular routing results before making activation decisions.
- Rerun optimization checks after major changes to queues, teams, skills, flows, staffing, or outcome definitions.
- Use authorized customer and interaction data only, and don't document real tokens or customer identifiers in API examples.
Limits and considerations
Configuration limits
| Limit | Value |
|---|---|
| Maximum queues per organization that can use AI routing | 100 queues |
| Maximum queues that can run evaluation mode at the same time per organization | 100 queues |
| Maximum queues in an optimization check request | 50 queue IDs |
| Maximum queues in an optimization status request | 100 queue IDs |
Routing considerations
- AI routing doesn't override queue ranking, contact priority, skills, teams, channel capacity, availability, or flow constraints.
- New queues and organizations need enough historical interaction data before AI routing can be evaluated effectively.
- Evaluation and reporting include only interactions longer than 30 seconds.
- AI routing uses up to 90 days of historical interaction data for model training.
- Custom outcomes must be captured at the interaction level and must remain consistent over time.
- Use evaluation mode before activating AI routing for live traffic.
Troubleshooting
| Issue | What to do |
|---|---|
| The queue shows low potential. | Continue regular routing. Review the selected outcome, queue volume, historical data, and agent distribution, then run the optimization check again when more data is available. |
| The optimization check doesn't start. | Confirm that AI routing is enabled for the organization, the selected queues are valid, and your role has permission to configure routing. |
| Evaluation mode doesn't show results immediately. | For first-time evaluation, model training happens during regional off-hours. Review results the next day after new interactions arrive. |
| The evaluation report has fewer interactions than expected. | Confirm that the queue received eligible traffic. Interactions that are 30 seconds or shorter aren't included in AI routing reports. |
| Activation isn't available. | Confirm that AI routing is enabled for the organization, evaluation mode has produced enough data, and the organization has the required entitlement. A low-potential optimization result is advisory and doesn't prevent activation. |
| A custom KPI isn't available. | Confirm that the global variable or custom outcome is configured, populated consistently, and available as an interaction-level value. |
| AI routing doesn't improve the selected outcome. | Review the selected KPI, data quality, queue setup, staffing patterns, and recent operational changes. Consider running evaluation again before continuing live use. |
