Over the past years, Robotic Process Automation (RPA) has emerged as a significant tool to enhance productivity across various industries by automating repetitive tasks performed on computer user interfaces, thereby reducing error rates. In this paper, the RPA problem is addressed as an unbounded assignment problem. Specifically, the objective of the problem is to assign robots to transactions that must be completed at specific time periods, while minimizing the total number of robots required. The problem is solved using a bipartite graph representation and a random walk approximation, which defines an ordering of the transactions and periods in order to determine a valid assignment. The heuristic is evaluated on a real data set from a financial institution and compared to previous results obtained on generated data. The results obtained with the random walk approximation heuristics on the real data set is optimal in terms of number of robots.