Policy institutions often address uncertainty around a baseline outlook through a limited set of alternative scenarios, rather than specifying a complete predictive distribution. This paper examines optimal policy prescriptions within this setting with a focus on the effect of uncertainty about the outlook. We propose a framework where policymakers form beliefs about the likelihood of a baseline economic path and alternative scenarios, updating these beliefs over time using Bayes’ rule. This “risk-adjusted optimal policy” allows policymakers to set policy rationally while learning which scenario is unfolding. We apply this framework to the U.S. economy at the end of 2024, considering risks of higher inflation and recession around a baseline projection.