Rethinking manufacturing agility assessment in the age of intelligent systems: a multi-agent skill-enhanced RAG framework for the digital manufacturing era

Ren, J ORCID: 0000-0003-4640-824X (2026) Rethinking manufacturing agility assessment in the age of intelligent systems: a multi-agent skill-enhanced RAG framework for the digital manufacturing era. The International Journal of Advanced Manufacturing Technology. ISSN 0268-3768

Preview

Text s00170-026-18700-7.pdf - Published Version Available under License Creative Commons Attribution. Download (1MB) | Preview

Abstract

Manufacturing agility has been a strategic imperative since its formalisation at the Iacocca Institute in 1991. Over three decades, researchers have developed attribute-based frameworks that identify the organisational, technological, and relational dimensions of an agile enterprise. Yet the methods used to assess those dimensions have remained largely unchanged. Data collection still relies on surveys and site visits. Weights are still estimated through expert pairwise comparison. Synthesis still produces a single, static report. This paper argues that these methods are no longer adequate for the volatile, digitally interconnected manufacturing landscape of the mid-2020s. The response is not to apply a new computational tool to an old assessment framework. That approach merely replaces the surface while preserving the underlying limitations. Instead, this paper proposes an evolved framework with a new architecture: a Multi-Agent Skill-Enhanced Retrieval-Augmented Generation (RAG) system in which agility knowledge is encoded in versioned, learnable Skill modules, knowledge acquisition is automated through grounded document retrieval, weight calibration is performed by large language model (LLM) agents, and synthesis preserves formal Dempster-Shafer evidential reasoning throughout. Four agility domains are selected and updated for the Industry 4.0 era: Technology, Partnership, Market, and Change. Each is extended with a new attribute not captured in prior frameworks. The result is a system that learns continuously, remains auditable, and is deployable by manufacturing firms of all sizes. Small and medium enterprises (SMEs) stand to benefit most.

Item Type:	Article
Uncontrolled Keywords:	01 Mathematical Sciences; 08 Information and Computing Sciences; 09 Engineering; Industrial Engineering & Automation; 40 Engineering; 46 Information and computing sciences; 49 Mathematical sciences
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) T Technology > TS Manufactures
Divisions:	Engineering
Publisher:	Springer
Date of acceptance:	5 July 2026
Date of first compliant Open Access:	10 July 2026
Date Deposited:	10 Jul 2026 11:42
Last Modified:	10 Jul 2026 11:42
DOI or ID number:	10.1007/s00170-026-18700-7
URI:	https://researchonline.ljmu.ac.uk/id/eprint/28990

View Item