If the voting doesn't work it should be possible to gather other metrics such as how long the message was visible on the screen - in other words did the recipient actually read the email. Over time it would learn the users typical reading speed. Another would be how long was it sitting in the inbox before it was read. A smart email client can detect when the recipient actually notices that the email is there.

With these metrics it could then work out which members of the company send out the least well received emails. There are certainly some people where I work who have a very low interesting email ratio. These people would be prompted to sort this out.

One risk would be the employee that just sends out jokes would of course appear to be highly effective. Some algorithms would be needed to filter out work vs play emails.

To creep employees out, these metrics could be readily available to management and used in their performance reviews as part of all the other data obtained about them!