Two main trends characterize today's communication landscape and are finding their way into industrial facilities: the rollout of 5G with its distinct support for vertical industries and the increasing success of machine learning (ML). The combination of those two technologies open the doors to many exciting industrial applications and its impact is expected to rapidly increase in the coming years, given the abundant data growth and the availability of powerful edge computers in production facilities. Unlike most previous work that has considered the application of 5G and ML in industrial environment separately, this paper highlights the potential and synergies that result from combining them. The overall vision presented here generates from the KICK project, a collaboration of several partners from the manufacturing and communication industry as well as research institutes. This unprecedented blend of 5G and ML expertise creates a unique perspective on ML-supported industrial communications and their role in facilitating industrial automation. The paper identifies key open industrial challenges that are grouped into four use cases: wireless connectivity and edge-cloud integration, flexibility in network reconfiguration, dynamicity of heterogeneous network services, and mobility of robots and vehicles. Moreover, the paper provides insights into the advantages of ML-based industrial communications and discusses current challenges of data acquisition in real systems.