Development of this paper was proposed during discussions of the NAM Clinical Effectiveness Research Innovation Collaborative, in which several of the authors are participants. Due to our affiliation, we use a learning health system perspective, especially as it relates to clinical effectiveness research and practice. This paper focuses on a subset of key principles foundational to continuous learning. We revisit the distinctions between activities that we believe should require formal oversight by an IRB and those that should not. We provide more in-depth discussion on two aspects of the learning health system: the oversight of QI activities, and the use of cluster randomization as a tool to advance learning, both for QI and other operations activities. We focus especially on establishing an IRB oversight mechanism that protects individuals while encouraging health care systems to learn as much as possible from their ongoing activities.