I am currently analyzing the impact of AI-driven routing on our ‘Average Handle Time’ (AHT) and ‘Customer Satisfaction’ (CSAT) scores. I am seeing a slight improvement in AHT, but our CSAT scores have actually dropped in some queues since we enabled predictive routing. I suspect the AI is prioritizing speed over agent proficiency for specific call types. How can I pull the ‘Predictive Routing’ metadata via the Analytics API to see exactly which agents were selected by the AI and what the predicted KPI impact was for each call?
I deal with the voice quality for our remote agents. I have seen predictive routing cause issues when the AI selects an agent with a poor network connection just because they have the best ‘predicted’ handle time. Har40, you can find the predictive routing data in the v2.analytics.conversations.{id}.details endpoint. Look for the routingData object within the conversation record. It should contain the ‘Scoring’ details and the ‘Agent Scores’ that the AI used to make the decision.
I have been trying to put this predictive data into our custom WebSocket monitor. Har40, be aware that the routingData is only available after the call has been routed. If you want to see the predictions in real time, you have to use the Notification API to subscribe to the conversation events. It is a lot of data to process, but it is the only way to see the ‘AI Logic’ as it happens.
Hello again! To follow up on Roh14, please remember that predictive routing is a ‘Long-Term’ game. You should not judge the performance based on a few days of data. The AI needs time to learn from the agent outcomes. I always recommend running an ‘A/B Test’ for at least thirty days before making any final decisions on the KPI impacts.