Authors: Stephen Hudson (Argonne National Laboratory), Kyle Chard (University of Chicago, Argonne National Laboratory), Rafael Ferreira da Silva (Oak Ridge National Laboratory (ORNL)), Daniela Cassol (Lawrence Berkeley National Laboratory), Woong Shin (Oak Ridge National Laboratory), Mathew Cherukara (Argonne National Laboratory), Mathew Sinclair (Argonne National Laboratory), Logan Ward (NVIDIA)
Abstract: This BoF will convene the workflows community to discuss the challenges of next-generation scientific workflows. A central focus will be on how the role of the human changes as workflows become increasingly autonomous: where people remain in the loop, how they steer and validate agentic systems, and how our community can support workflow developers.
The session will bring together workflow users, developers, theorists, and industrial collaborators, highlighting emerging research directions and practical needs from scientific campaigns. Through interactive discussions, participants will explore collaborative strategies to enable resilient, scalable, and adaptive workflows that meet the evolving demands of scientific discovery.
Long Description: Scientific workflows are entering a new phase in which AI, agentic systems, and cloud-service-driven infrastructure are reshaping how scientific campaigns are executed, steered, and validated. Increasingly, workflows are becoming the front end to HPC platforms, data repositories, models, inference services, and experimental facilities, creating new challenges for coordination, autonomy, and human oversight. This BoF builds on four prior successful sessions at SC22, SC23, SC24, and SC25, which have each gathered between 85 and 120 participants.
This year, we focus the conversation on how the rapid rise of agentic workflows is reshaping scientific workflow execution. First, we will examine how AI agents are changing workflows from predefined execution plans into adaptive systems that can reason, plan, and act during a campaign. Second, we will explore what this means for the human role: how scientists and workflow developers steer autonomous behavior, intervene when needed, and validate outcomes. Third, we will discuss the new execution landscape these systems require, where workflows connect cloud services, HPC, data, models, inference endpoints, and experimental facilities into a coordinated environment. Finally, we will consider the community practices, infrastructure, and policies needed to make these systems reliable, scalable, and usable for real scientific campaigns.
This BoF will bring together researchers, facility operators, and workflow developers to assess the current landscape and envision the next phase of workflow-enabled science. Participants will hear from experts who are leading national and international initiatives in agentic workflows, HPC/AI integration, and cross-facility execution. The session will feature lightning talks to set the stage, followed by a highly interactive format where attendees will engage in group discussions focused on identifying key challenges and actionable next steps.
Outcomes from the session will include a summary of community-identified barriers and priorities, as well as an invitation to contribute to future collaborative efforts within the Workflows Community Initiative. We aim to catalyze new collaborations, identify policy and infrastructure gaps, and chart a community-informed roadmap for workflow innovation.
SC26 provides an ideal venue for this discussion by bringing together a global community of stakeholders in high-performance and data-intensive computing. This BoF will serve as a collaborative forum to define the requirements for resilient, intelligent, and coordinated workflows that can meet the demands of the next generation of scientific discovery.
Website: https://workflows.community/bof/sc26