With increasing demand for customized connectivity, transport networks must evolve towards autonomous and customer-driven network management. This paper presents a comprehensive overview of network autonomy and the challenges associated with evolving toward higher levels of autonomy. Moreover, various use cases of artificial intelligence in network automation in IP-over-DWDM transport networks are also analyzed, in particular related to traffic prediction, quality of transmission, anomaly detection, network planning, and proactive failure management. Additionally, the role of generative AI in network operation is explored. Central to our discussion is a proposed control architecture based on open and standard SDN APIs, which incorporates network slicing for multi-layer transport networks and enables real-time access to normalized data, facilitating autonomous network operation.
