The current Internet-of-Things (IoT) networks in 5G new radio (NR) are bandwidth hungry to maintain the resource utilization demands. Hence, unlicensed spectrum sharing has evolved as a critical use case for IoT-NR networks. However, IoT-NR devices come with diverse features and their coexistence in the unlicensed spectrum can degrade the performance of its primary networks. Therefore, an efficient coexistence mechanism based on the idea of adaptive initial sensing duration (ISD) is proposed to enhance the IoT-NR network performance while keeping the primary Wi-Fi network's performance to a bearable threshold. A Q-learning (QL) based algorithm is devised to maximize the normalized sum throughput of the coexistence Wi-Fi/IoT-NR network. The results confirm a maximum throughput gain of 51 % and also ensure that the Wi-Fi network's performance remains intact.