Artificial intelligence adoption in public organisations: evidence from the Italian welfare agency

Abstract

Purpose – This study aims to examine how organisational, technological and environmental factors interact to enable systematic artificial intelligence (AI) adoption in a large welfare administration operating under constitutional social protection obligations and demographic pressure; it further asks whether, and how, these factors manifest differently across two categories of AI use (citizen-facing versus internal administrative systems). Design/methodology/approach – Using a single-case study of Italy's National Social Security Institute, the research applies an adapted technology-organisation-environment (TOE) framework for the public sector. The empirical base combines 10 semi-structured interviews with senior executives, extensive analysis of regulatory and strategic documents and validated performance indicators for a broad AI portfolio. Findings – Environmental pressures – ageing populations, constitutional obligations and fiscal constraints – are the main drivers of AI adoption, altering the relative weight of dimensions in the traditional TOE framework. Systematic implementation is sustained by performance measurement and monitoring, robust data governance on a mature technological infrastructure and leadership that explicitly embeds AI within the social protection mandate. Research limitations/implications – As a single-case study focused on senior leadership, the findings require comparative validation across institutional and national settings. Transferability is conditioned on the presence of constitutional welfare mandates, strong demographic pressure and mature information technology infrastructure. Practical implications – AI adoption should follow a clearly defined strategy, supported by performance systems tracking efficiency and equity and by leadership able to translate constitutional principles into technological governance. Originality/value – The study addresses the lack of empirical research on AI adoption in large European welfare institutions beyond Anglo-Saxon and Nordic contexts. We propose organisational metabolism as a mid-range theoretical construct, a sector-specific refinement of dynamic capabilities, denoting a sustained organisational capacity for continuous technological absorption, adaptive reconfiguration and normative calibration to constitutional values across political and regulatory cycles.