Despite our familiarity with the concept of habits, eliciting and measuring habits experimentally in humans has proven to be difficult. A possible explanation is that participants in psychological experiments actively recruit goal-directed control and therefore make few habitual slips-of-action in the presence of stimuli signalling devalued outcomes. In the current experiment we used the symmetrical outcome-revaluation task in combination with a working memory load in an attempt to tip the balance from goal-directed control to stimulus-response habit. During the instrumental learning phase, participants learned to make a Go response to stimuli signalling valuable outcomes (and points) while not responding (NoGo) to stimuli signalling not-valuable outcomes. During the subsequent test phase, the outcomes signalled by the stimuli were either value-congruent with training (still-valuable and still-not-valuable), or value-incongruent (devalued and upvalued). Participants had to flexibly adjust their behaviour on value-incongruent trials where the stimulus-response association learned during training was no longer appropriate. For half the participants, a concurrent working memory load was imposed during the test phase. In line with our preregistered hypotheses, participants showed evidence for habitual slips-of-action but those under working memory load showed increased habit tendencies (specifically failures to inhibit prepotent Go responses in the presence of stimuli signalling devalued outcomes). This central finding suggests that a working memory load can be used to reveal habits in humans.