Future cellular networks will support a massive number of devices as a result of emerging technologies such as Internet-of-Things and sensor networks. Enhanced by machine type communication (MTC), low-power low-complex devices in the order of billions are projected to receive service from cellular networks. Contrary to traditional networks which are designed to handle human driven traffic, future networks must cope with MTC based systems that exhibit sparse traffic properties, operate with small packets and contain a large number of devices. Such a system requires smarter control signaling schemes for efficient use of system resources. In this work, we consider a grant-free random access cellular network and propose an approach which jointly detects user activity and single information bit per packet. The proposed approach is inspired by the approximate message passing (AMP) and demonstrates a superior performance compared to the original AMP approach. Furthermore, the numerical analysis reveals that the performance of the proposed approach scales with number of devices, which makes it suitable for user detection in cellular networks with massive number of devices.