Computational workflows have become central to modern data-driven scientific discovery, underpinning analytical pipelines across domains from genomics and climate science to high-energy physics and biodiversity research. Despite this ubiquity, workflows remain poorly described and difficult to find or reuse outside their original context. This seminar addresses a foundational challenge: establishing the principles, standards, and infrastructure needed to make computational workflows FAIR (Findable, Accessible, Interoperable, and Reusable) at a time when governments, funders, and research infrastructures worldwide expect research artefacts (software, data, workflows) to be FAIR.

A distinctive challenge in making workflows FAIR is their dual nature. As epistemic objects, workflows document scientific methods and carry provenance records of how results were produced; as executable objects, they carry the characteristic demands of software: versioning, portability, dependency management, and long-term maintainability. FAIR frameworks developed primarily for datasets cannot be applied directly to this dual nature, and workflows must also be assessed against the emerging CODE beyond FAIR movement, which emphasizes that research code must be Open, Documented, Executable, and Collaborative. Several concrete obstacles have impeded progress: workflow metadata is rarely machine-actionable and remains tightly coupled to individual management systems; no agreed standards exist for identifier granularity or provenance capture at scale; and incentives to share reusable workflows remain weak across most research communities.

Compounding these challenges is the fragmentation of the communities involved. Workflow researchers, developers of widely-used workflow systems such as Galaxy, Nextflow, Snakemake, and Common Workflow Language (CWL), domain scientists, research infrastructure operators, and the scholarly communication community all have a stake in FAIR workflows, yet rarely occupy the same forum. The infrastructure for progress (WorkflowHub, Dockstore, Software Heritage, RO-Crate, CodeMeta, and the DataCite PID Graph) already exists in nascent form. What is missing is consensus on the standards, semantics, and minimal requirements for coherent interoperability.

This seminar aims to bring together stakeholders from across the workflow, FAIR, and research community to address these challenges. Representatives of different disciplines, popular workflow platforms and associated FAIR services, commercial and open-source providers, research practitioners, digital Research Technical Professionals, AI-enabled software development, policymakers and funders will focus on five interconnected topics essential to realizing FAIR workflows in practice: (1) the machine-actionable metadata needed for workflows to be discoverable and interpretable across systems; (2) persistent identifier (PID) schemes to support citation, attribution, and tracking of workflow variants and reuse; (3) provenance models for reproducible and comparable workflow runs across configurations and environments; (4) the services (registries, repositories, execution platforms, and monitoring tools) required to sustain a FAIR workflow ecosystem; and (5) the practical strategies for supporting adoption across communities with heterogeneous tooling and varying maturity. Common to all five topics is the role of generative AI, which offers new possibilities for extracting and enriching workflow metadata, identifying variants, and augmenting attribution graphs, whilst raising questions about the reliability of automatically generated provenance records.

This Dagstuhl Seminar is expected to build that consensus, with the aim of producing a community roadmap with concrete recommendations for workflow systems, metadata standards, identifier schemes, and provenance models to shape how computational workflows are shared, cited, and sustained.

Creative Commons BY 4.0

Sean R. Wilkinson, Sarah Cohen-Boulakia, Anna-Lena Lamprecht, and Carole Goble