I would be outraged IF I found out. If the AI-to-human hand-off is smooth enough, there is no way to figure this out. In your scenario, if the AI's only task is to send gazillions of emails to generate leads, and then the human takes over when the leads come in, the respondents have no way to figure out that the initial email was an AI.

Of course, the answer is to have AI send a response with a CAPTCHA (assuming those still work), before showing the initial email to the recipient.