The proper optimization of business processes remains vital because it leads to more efficient operations as well as less costs and higher task completion rates. The researchers use Q-Learning from the Reinforcement Learning framework to improve workflow decisions made by simulated agents. The processing of 500 entries included standardization and discretization techniques that normalized distributions after the Shapiro-Wilk test proved non-normal data. The agent operated with an ε-greedy strategy while updating its Q-values based on Bellman’s equation in the 41-dimensional state space feature vector. The agent initially performed 1,000 trials which resulted in achieving 45.08 average reward when picking “Allocate Resource” as the primary choice. The agent’s performance increased by 3.48% after hyperparameter optimization when learning rate was reduced, discount factor was increased, and exploration decay was slowed down while achieving 46.65 average reward during 500 trials. The research results demonstrate that workflow efficiency benefits from adaptive RL methods which automatically optimize how tasks get performed. Research about RL conducted by Sutton and Barto (2018) [1] is closely aligned with this work which reinforces the benefits that reward-driven learning brings together with balanced exploration-exploitation balances.