In recent years, human workers in manual assembly lines are increasingly being supported by the deployment of complementary technology. Collaborative robots (or cobots) represent a low-threshold opportunity for partial automation and are increasingly being utilised by manufacturing corporations. As collaborative robots can be used to either conduct tasks in parallel to the human worker or collaborate with the worker on an identic task, industrial planners experience an increasingly complex environment of assembly line balancing. This contribution proposes three different decomposition approaches for Benders’ decomposition algorithms exploring the multiple possible partitions of the formulation variables. We evaluate the performance of the algorithms by conducting extensive computational experiments using test instances from literature and compare the findings with results generated by a commercial solver and a metaheuristic solution procedure. The results demonstrate the Benders’ decomposition algorithms’ efficiency of finding exact solutions even for large instances, outperforming the benchmark procedures in computational effort and solution quality.