This paper uses Q-learning algorithm to train agents in a R&D competition environment, and discusses the possible results caused by artificial intelligence decision. The results show that: first, it is common for lower R&D investment and higher profit levels than Nash equilibrium results. Second, the increase in the number of agents can reduce the probability of achieving “collusion”. Third, the spillover degree of R&D will promote the “collusion” of agents, and it is easier to achieve when the number of agents is large. Fourth, the increase in the learning rate can reduce the probability of achieving “collusion”, and it is more obvious when the number of agents is large and the spillover of R&D is low. Based on the above research results, this paper puts forward three suggestions to strengthen the management of artificial intelligence application at the decision-making level of enterprises, which provides theoretical support for the continuous improvement of anti-monopoly regulation in the era of digital economy.