Our goals sometimes conflict with our prepotent habitual responses, which often leads to impaired performance on a variety of tasks. People are better at exerting cognitive control to overcome prepotent and automatic responses when they are motivated by the prospect of reward. The standard experimental paradigms used to study this phenomenon examine free RTs that allow participants to select a variety of response strategies including delaying response initiation to avoid committing errors. However, this approach makes it difficult to determine which control processes are affected by reward. Does reward lead to improved performance via the inhibition of prepotent responses or the facilitation of goal-directed processing? Here, we use a forced-response paradigm to fix response initiation and systematically vary the time available for the cognitive processing necessary for response preparation. Using a probabilistic model that dissociates the preparation of habitual and goal-directed responses, we obtain evidence across multiple experiments ( n = 87 people) that reward selectively accelerates the preparation of goal-directed actions in the context of conflict.