Opus: A Workflow Intention Framework for Complex Workflow Generation
Phillip Kingston, Théo Fagnoni, Mahsun Altin
TL;DR
Opus presents a formal framework for extracting and encoding process objectives from Business Artefacts to enable AI-assisted, AI-generated workflows. It introduces Workflow Signals as vectors and Workflow Intention as a tensor, defined by the triple $(i,p,o)$ that aligns Input, Process, and Output across modalities. The approach employs a modular, attention-based multimodal architecture with modality-specific encoders, intra-modality attention, inter-modality fusion, and an intention decoder, together with a two-stage training procedure and specialized loss functions. The framework addresses practical concerns of scalability and applicability in complex business environments, offering a pathway to rapid supervised automation and AI-enhanced workflow evolution.
Abstract
This paper introduces Workflow Intention, a novel framework for identifying and encoding process objectives within complex business environments. Workflow Intention is the alignment of Input, Process and Output elements defining a Workflow's transformation objective interpreted from Workflow Signal inside Business Artefacts. It specifies how Input is processed to achieve desired Output, incorporating quality standards, business rules, compliance requirements and constraints. We adopt an end-to-end Business Artefact Encoder and Workflow Signal interpretation methodology involving four steps: Modality-Specific Encoding, Intra-Modality Attention, Inter-Modality Fusion Attention then Intention Decoding. We provide training procedures and critical loss function definitions. In this paper we introduce the concepts of Workflow Signal and Workflow Intention, where Workflow Signal decomposed into Input, Process and Output elements is interpreted from Business Artefacts, and Workflow Intention is a complete triple of these elements. We introduce a mathematical framework for representing Workflow Signal as a vector and Workflow Intention as a tensor, formalizing properties of these objects. Finally, we propose a modular, scalable, trainable, attention-based multimodal generative system to resolve Workflow Intention from Business Artefacts.