In a wireless communication network, the propagation medium has been perceived as a randomly behaving entity between the transmitter and receiver for a long time. However future wireless networks, such as 5G and beyond, rely on Intelligent Reflecting Surfaces (IRS) for promising energy efficiency and spectrum improvement as it enables the proactive control of the wireless channel. In this article, we take a step further and discuss the role of Deep Reinforcement Learning (DRL)-based IRS-assisted network in delivering performance enhancement in a wireless communication network. We provide an insight into DRL followed by the challenges in place for an IRS-assisted wireless network. We then lay out a case study to show the channel processing using two state-of-the-art DRL algorithms, that is, deep Q-network (DQN) and deep deterministic policy gradient (DDPG), in an IRS-assisted network and work on jointly optimizing the transmit beamforming and the IRS phase angles for sum-rate maximization. The sum-rate improvement is shown in reference to the number of reflecting IRS elements. Opportunities in DRL-based IRS-assisted wireless networks are then discussed to showcase the exciting future opportunities in this domain.