From Automation to Agentic Intelligence: A Framework for AI-Driven Digital Transformation, Innovation Capability, and Strategic Decision-Making in Contemporary Enterprises
TL;DR
The authors present a conceptual framework for the relationship between AI capability inputs and organizational absorption processes and downstream effects, which are moderated by the regulatory environment and innovation industry context and mediated by innovation capability and employee AI literacy.
Abstract
From descriptive analytics to generative systems and, most recently, autonomous agentic architectures, artificial intelligence (AI) is transforming how enterprises achieve digital transformation and develop innovation capabilities and how they shape strategic decision-making. This study highlights an AI adoption-value paradox; despite significant investments, many organizations report a lack of business value from AI. The authors present a conceptual framework for the relationship between AI capability inputs (generative, predictive, and agentic) and organizational absorption processes (culture, skills, and governance) and downstream effects (digital transformation maturity, innovation performance, and decision quality), which are moderated by the regulatory environment and innovation industry context and mediated by innovation capability and employee AI literacy. Based on decision intelligence theory, the technology-organization-environment (TOE) framework, and dynamic capabilities theory, this study follows a mixed-methods approach with a structured survey of mid-to large-sized enterprises, followed by semi-structured interviews with C-suite and senior IT executives that will be analyzed through partial least squares structural equation modelling (PLSEM) and thematic analysis. This study extends the dynamic capabilities theory to the domain of agentic AI, provides managers with practical guidelines for governing AI decisions, and provides regulators with insights into responsible AI adoption. The implications, limitations, and directions for empirical validation are discussed.