Previous research in reverse auction B2B exchanges found that in an environment where sellers collectively can cater to the total demand, with the final (i.e. the highest-priced bidding) seller catering to a residual, the sellers resort to a mixed strategy equilibrium [2]. While price randomization in industrial bids is an accepted norm, it may be argued that managers in reality do not resort to advanced game theoretic calculations to bid for an order. What is more likely is that managers learn that strategy and over time finally converge towards the theoretic equilibrium. To test this assertion, we model the two-player game in a synthetic environment, where the agents use a simple reinforcement learning algorithm to put progressively more weights on selecting price bands where they make higher profits. We find that after a sufficient number of iterations, the agents do indeed converge towards the theoretic equilibrium.