Cognitive impairments in bipolar disorder (BD) are prevalent but effective treatments with replicated and lasting pro-cognitive effects are lacking. Treatment development is hampered by a lack of neurocircuitry biomarkers to predict treatment efficacy. Action-Based Cognitive Remediation (ABCR) improves executive function in BD and this was accompanied by increased dorsal prefrontal cortex (dPFC) response during working memory (WM) after two weeks of treatment. This study investigated whether pre-treatment WM-related dPFC response, executive dysfunction and/or subjective cognitive difficulties predicted ABCR treatment response on executive functions. Forty-five patients with fully or partially remitted BD (ABCR: n = 25, control treatment: n = 20) in our ABCR trial completed a spatial N-back WM task during functional magnetic resonance imaging (fMRI) at baseline. Patients also completed neuropsychological tests and rated their cognitive functions before and after 10 weeks of ABCR or control treatment. Multiple linear regression analyses were conducted to assess whether pre-treatment dPFC response, objective executive impairment and/or subjective cognitive difficulties predicted greater ABCR-related improvements of executive function. We found that treatment-related improvement in executive function was predicted by more WM-related dPFC hypo-activity at baseline (p = 0.03) in linear regression analyses adjusted for age, gender and education. In contrast, there was only a non-significant trend towards more executive dysfunction at baseline predicting greater ABCR-related executive improvement (p = 0.08). Subjective cognitive difficulties at baseline showed no association with treatment effects (p = 0.16). In conclusion, pre-treatment dPFC hypo-activity during WM performance predicts greater effects of ABCR treatment on executive function and may represent a neurocircuitry biomarker for treatment efficacy in this cognitive domain.