The integration of artificial intelligence as a foundational enterprise system has created a profound structural tension between the need for centralised governance and the imperative for decentralised agility. This analysis explores the "centralisation-agility trade-off," arguing that true organisational resilience in the AI era requires a sophisticated redesign of enterprise architecture rather than a choice between control and innovation. Excessive centralisation, while ensuring compliance and mitigating risk, inevitably creates operational bottlenecks and, ironically, fosters the dangerous proliferation of "Shadow AI"—unsanctioned tool use that exposes firms to catastrophic data leaks and intellectual property theft. Conversely, unchecked decentralisation leads to chaos and an inability to scale securely. The resolution lies in a federated model that strategically combines central stability with peripheral flexibility. This involves evolving the operating model from a restrictive Centre of Excellence (CoE) to an empowering Centre for Enablement (C4E), which uses platform engineering to provide secure, self-service AI tools. At the data layer, a hybrid architecture synthesising the automated intelligence of a Data Fabric with the domain-driven ownership of a Data Mesh provides a robust foundation. Technologically, advancements such as Federated Learning enable collaborative model training on decentralised data without compromising privacy. These architectural shifts must be underpinned by adaptive governance frameworks that enable rapid, safe innovation. Ultimately, sustainable agility is not suppressed by standardisation but is built upon it, creating an environment of frictionless enablement where the most secure and compliant path for AI adoption is also the fastest and easiest for the end-user.
Partha Majumdar (Thu,) studied this question.