When we reason about others’ behavior, there are often many equally-plausible explanations. If Bob climbs a tree to get an apple, we may be unsure if Bob found climbing difficult but really wanted an apple; if he found climbing easy and was not particularly excited about the apple; or if he found climbing intrinsically fun and just got the apple because it was convenient. Past research suggests that we solve this problem by obtaining repeated observations about the agent and about the world. Here we argue that, beyond allowing us to sharpen our inferences about agents and the world, covariation data also enables us to do one-shot inferences about novel agents. We show that given minimal covariation data, people can infer objective and subjective properties of a new agent from a single event. We show that a model that assumes that agents maximize utilities matches participant judgments with quantitative precision.