Chronic alcohol abuse is frequently considered a habitual or inflexible behavior; however, measures of pre-existing cognitive flexibility prior to initiation of alcohol use are usually not available. This study used rhesus monkeys and an attentional set-shifting task to investigate whether pre-existing cognitive flexibility would predict increased risk for heavy alcohol drinking. As previously reported, monkeys were given 30 daily set-shifting sessions prior to alcohol access. These sessions consisted of the same sequence of eight unique visual discriminations (sets) of two objects that varied on two dimensions (shapes and colors). The ratio of errors per trials, session duration, and maximum set reached were primary dependent variables from each session and were used to compose a session performance index (PI) that ranged from a low performance PI of 31 to an optimal performance PI of 247. Here, animals underwent an alcohol induction period followed by 22 weeks of daily (22-h) self-administration sessions with free access to water and alcohol. Based on average daily alcohol intake during 22 weeks of 22-h/day access, the monkeys were categorized as non-heavy (mean = 2.0 ± 0.3 g/kg/day; n = 3) and heavy (mean = 3.3 ± 0.5 g/kg/day; n = 6) drinkers. The two groups diverged in performance on the set-shifting task across the 30 pre-alcohol sessions, and at the end of the pre-alcohol testing, the group average PI was 216 ± 27 and 137 ± 71 for the future non-heavy and heavy drinkers, respectively. The data show that low cognitive flexibility assessed with a set-shifting procedure was predictive of future classification as a heavy alcohol drinker. The data highlight individual differences in both cognitive flexibility and in alcohol self-administration in this population of rhesus monkeys.