When do you use bleeding edge technology?

When is the next big thing the right thing? Bleeding edge tech is a risk, but sometimes it is a risk worth taking. How do you approach bringing brand new tech into production?