Two different scenarios. That is the time from processing your speech, understanding it, taking action and give response. Its not necessary comparable to a scenario with traffic controlled by AI where actions may already be predicted and multiple predictions/actions are batch processed.
Also response time has be improving a lot in the beta period of SayI and chatGPT will essentially come with faster models.