Pipelines of computation to extract insights from raw data continue to increase in complexity and scale. When input data sources and output-consuming applications span globally distributed networks, establishing and maintaining these pipelines becomes particularly challenging. This work automates the joint planning and scheduling of data-intensive workflows, permanently deployed pipelines across cloud, fog, and edge sites, in a vendor- and platform-agnostic manner. I propose a network services model introducing two common network-level scheduling targets: data sharing and data processing interfaces. These interfaces are analogous to the Session and Presentation layers of the Open Systems Interconnect Model. I develop a formal workflow and resource graph formulation over these interfaces that explicitly decouples processing placement, data placement, and routing decisions. Leveraging this formulation, I introduce the WORKSWORLD framework and three planning domains. The WORKSWORLD and WORKSWORLD-2 domains encode the problem for linear-chain and series-parallel workflows for numeric planning; WORKSWORLD-2H for hierarchical task network planning. The framework accepts user-defined declarative configuration files (i.e. YAML) specifying data sources, available workflow components, and desired data destinations and formats, translates them into planning problems, invokes domain-independent AI planners, and validates and visualizes the resulting workflow graph and schedule. I benchmark a state-of-the-art numeric planner and a hierarchical planner against my publicly released benchmark suite, demonstrating that each planner occupies a distinct point on the scalability–quality tradeoff. The numeric planner produces cost-efficient plans on moderate-sized resource and workflow graphs on commodity hardware, while the hierarchical planner solves substantially larger instances in near-real time at a modest increase in plan cost. These results ground a theoretical analysis motivating hierarchical and multi-agent planning architectures for collaborative workflow scheduling at scale. My results suggest a new network model and modern AI planning techniques can support orchestration of universal workflows across hybrid and multi-cloud enterprises. This grounds my discussion of future work to improve my planning domains, prototype data sharing and processing interfaces and identify open protocols to expose them to all network devices exchanging pipelined data.