I am currently monitoring the ‘Agent Matching’ latency for our new AI-driven routing model. We are seeing that some interactions are sitting in the queue for several extra seconds while the ‘Predictive Engine’ calculates the optimal agent match. Is there a way to tune the ‘Scoring Threshold’ to reduce this latency, or is this delay a standard trade-off for the improved matching accuracy of the predictive model?
I have been building a custom dashboard to track these routing latencies. Fat85, the delay you are seeing is usually related to the ‘Data Density’ of your model. If the engine has to process a huge number of agent and customer attributes, the scoring will take longer. You should check if you are sending unnecessary ‘Custom Attributes’ to the model. Reducing the number of input features can significantly speed up the calculation without sacrificing too much accuracy.
I have seen these routing delays impact our ‘First Response Time’ metrics in Salesforce. To follow up on Isa30, you should also look at your ‘Bullseye Routing’ fallback. If the predictive model fails to find a match within your ‘Timeout’ window, it will fall back to standard routing. If your timeout is set too long, the customer is just sitting there waiting for the AI to give up. We found that a five-second predictive timeout was the ‘Sweet Spot’ for our high-volume support queues.
Hello everyone! I love seeing how AI is improving the contact center experience. Fat85, one more thing to consider: the predictive engine performs better when it has a ‘Large Pool’ of available agents to choose from. If your queue only has two or three agents, the engine might struggle to find a significant difference in their scores, which can add to the processing time. Predictive routing is most effective (and fastest) for queues with at least twenty concurrent agents!