Generative AI for Enterprise Workflow Automation: Scalability, Governance, and Performance

Abstract

Enterprises have automated business processes for decades using business-process-management (BPM) notations and, more recently, robotic process automation (RPA), but both approaches require every step of a workflow to be explicitly modeled in advance, leaving them brittle whenever a process encounters a case its designers did not anticipate. Generative AI, and specifically large-language-model agents capable of maintaining memory across multi-step tasks, offer a way to handle exactly these unanticipated cases, but introducing autonomous, generative decision-making into a workflow that was previously fully deterministic raises governance and scalability concerns that neither the BPM nor the RPA literature was built to address: an agent empowered to take unsupervised action inside an enterprise process needs a governance mechanism deciding when its action may proceed autonomously and when it must be reviewed by a person, and the infrastructure running that agent needs to serve highly variable, bursty inference load without degrading the latency of the business process it supports. This paper proposes GOWA (Generative-AI Orchestrated Workflow Automation), an architecture in which an agentic workflow orchestrator routes BPMN-style process steps through memory-aware large-language-model agents, every agent action is passed through a risk-tiering governance gate aligned with the NIST AI Risk Management Framework before it is allowed to execute autonomously or escalated to a human reviewer, and the resulting workload is served by a scalability layer purpose-built for efficient large-language-model inference under container orchestration. Following a design-science research approach, GOWA is not empirically validated, through a comparative capability table, a short case scenario, and two illustrative figures, with every quantitative value explicitly labeled as an estimate rather than measured production data. The paper situates GOWA against the literature on business-process management and robotic process automation, generative-AI agent capabilities and memory mechanisms, responsible-AI governance frameworks, and scalable infrastructure for large-language-model serving.