There isn't currently a great way to do this if you previously had 100% of traffic in treatment and control. We're looking to support this use case more in the future.
For now, if starting from the beginning, you can experiment on 10% of traffic, (5% vs 5%), keep that experiment going, and then just change the default for other users from OFF to ON. Then the experiment stays live but you've released the treatment to 95% of users.