Generative AI is racing ahead of public-sector digital transformation research, major review finds

Generative artificial intelligence is spreading rapidly through government offices around the world, yet the scholarly literature on public-sector digital transformation has barely begun to grapple with it. That is the central conclusion of a systematic literature review published in Discover Artificial Intelligence, which synthesised 125 peer-reviewed studies published between 2021 and 2026 to map how researchers currently understand digital transformation in government in the era of AI. The review, conducted by Gideon Mekonnen Jonathan following the PRISMA 2020 reporting standard, finds that only a small subset of the literature substantively engages with AI at all, and that still fewer studies address generative AI, large language models or foundation models — a widening gap between what is happening in public administration and what scholars have documented about it.

The review set out to answer a broad question: how is digital transformation in the public sector being reshaped by the emergence of AI and, in particular, generative AI, and what factors, outcomes and challenges does the current scholarship identify? To do this, the author searched eight major databases, including Scopus, Web of Science, the AIS Electronic Library, ACM, IEEE Xplore, ScienceDirect, Wiley and Taylor & Francis, using a Boolean search string that combined digital-transformation terms, public-sector terms and AI terms ranging from machine learning to ChatGPT and Copilot. Inclusion required peer-reviewed journal articles or full conference papers in English published between January 2020 and March 2026, with all studies meeting the final criteria appearing between 2021 and 2026. Literature reviews, editorials, opinion pieces and private-sector studies were excluded.

Methodological quality was appraised with a four-item instrument scoring theoretical grounding, methodological transparency, sample and context description, and knowledge contribution on an ordinal scale from 0 to 8. Studies scoring at least 5 qualified for interpretive thematic synthesis; 88 of the 125 included studies, or 70.4 percent, met that threshold, while the remaining 37 were retained only for descriptive mapping. The corpus is dominated by qualitative case studies, complemented by structural equation modelling work and a smaller mixed-methods stream, while experimental and quasi-experimental designs remain rare. Europe is the dominant empirical setting, though the review notes significant representation from Africa, South-East Asia, the Middle East and Latin America.

Across the corpus, digital transformation is predominantly conceptualised as a sociotechnical and public-value-oriented process rather than the mere implementation of technology. The review identifies four interrelated conceptual strands: transformation as multidimensional organisational change involving culture, leadership, governance and service redesign; transformation evaluated through public value, spanning citizen, administrative, economic and societal dimensions; AI as an increasingly integral component of transformation, examined through decision-making, predictive analytics and human-AI collaboration; and transformation as co-produced through interactions among citizens, frontline officials and external actors. The author points to a study of citizen hacktivism in Stockholm as an illustration of how transformation processes can emerge from outside formal organisational structures.

Perhaps the most consistent finding concerns what drives transformation outcomes. Organisational and managerial factors — above all leadership, organisational culture and strategic alignment — emerge as the most reliable predictors of success or failure across highly diverse institutional contexts, from Scandinavian municipalities to Middle Eastern administrations, African public sectors and South-East Asian governance settings. Technological factors such as infrastructure, interoperability and legacy systems shape the pace, direction and unevenness of implementation, but the literature increasingly portrays technology as an enabling rather than determinative condition. Environmental factors, including regulatory frameworks, funding constraints, political conditions and crises such as the COVID-19 pandemic, are particularly prominent in studies from Global South contexts, where crises accelerated digitalisation while exposing underlying infrastructural and organisational weaknesses.

The review also identifies four explanatory dimensions that repeatedly appear across the corpus and seem increasingly difficult to subsume within the established organisational-technological-environmental framework. These are citizen co-production and participation; digitally induced administrative burden, including digital red tape and technostress; street-level administrative reconfiguration, in which digital systems redistribute rather than eliminate the discretionary authority of frontline officials; and multidimensional public-value evaluation, exemplified by a four-dimensional typology of citizen, administrative, economic and societal value. The author argues that the tripartite framework remains analytically robust but now requires extension to capture the organisational and normative complexities introduced by AI-enabled governance.

The opportunity profile of the literature is predominantly optimistic, with many studies associating digital transformation with increased productivity, cost reduction, service quality, administrative innovation, transparency, anti-corruption efforts and citizen access to services. Yet recurring risks are documented as well. Digital exclusion and inequality constitute the most consistently reported concern, followed by algorithmic bias, opacity and fairness issues, alongside privacy, cybersecurity and data-protection problems. The literature also flags digital overload, technostress and digital red tape as unintended organisational consequences, and political-economic critiques caution that transformation may facilitate new forms of surveillance, dependency and extractive governance.

What the literature largely fails to address, the review argues, are the risks specific to generative AI and foundation-model technologies. Sustained discussion of hallucination, prompt injection, training-data sovereignty and foundation-model governance remains largely absent, even though governments are already deploying generative systems in domains including welfare provision, healthcare, immigration and policing, often ahead of robust scholarly evaluation. The author identifies three conceptual challenges that existing scholarship has not adequately handled. First, foundation models are not fixed administrative tools but configurable generative systems whose outputs vary according to prompts, fine-tuning, training data and context, destabilising frameworks built around deterministic systems. Second, generative AI reshapes the organisational locus of transformation: whereas conventional digital transformation proceeds through centralised procurement and integration, generative tools are frequently adopted bottom-up by individual employees through subscriptions and browser-based tools, complicating procurement-centred governance models. Third, generative systems require continuous verification, oversight and contextual judgement, making hallucination management, prompt literacy, output reliability and model governance central administrative concerns.

The practical implications are pointed. Because leadership, organisational culture, strategic alignment and workforce competences consistently matter more than technological investment, the review suggests that governments should treat AI and generative AI deployment not as technology implementation projects but as broader organisational-transformation processes requiring sustained investment in leadership development, institutional learning and workforce adaptation. It further recommends that AI-enabled public services be evaluated within multidimensional public-value frameworks rather than through efficiency metrics alone, with transparency, accountability, participation, equity and democratic legitimacy built into assessment — a point the author deems especially important in high-consequence domains such as healthcare, welfare, policing and immigration, where algorithmic error or opacity can carry significant societal consequences.

The review is candid about its own limits. The search strategy was title-based and centred on digital-transformation terminology, which excluded a substantial body of adjacent AI-focused public-administration scholarship using vocabularies such as artificial intelligence, chatbots or foundation models; those studies informed contextual interpretation but were not part of the synthesis. Grey literature and policy reports were excluded, coding relied primarily on abstracts and conclusions, and screening, coding and quality appraisal were performed by a single reviewer without independent double-coding. Because the review is interpretive and qualitative, it cannot estimate effect sizes or causal magnitudes; conclusions about the importance of leadership or culture reflect consistency across studies rather than quantitative estimates of impact. With those caveats in place, the author calls for future research on generative AI governance, accountability and equity, on how generative tools reshape street-level administrative work, and on public-value and equity-oriented evaluation — arguing that the central task now is not to extend existing models incrementally but to build frameworks capable of explaining digitally mediated governance under conditions of generative automation.

Subject of Research: A systematic literature review of public-sector digital transformation in the era of AI and generative AI

Article Title: Public-sector digital transformation in the age of generative AI

Article References: Jonathan, G. M. (2026). Public-sector digital transformation in the age of generative AI. Discover Artificial Intelligence, 6(1), Article 968. https://doi.org/10.1007/s44163-026-02077-3

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02077-3

Keywords: generative AI, digital transformation, public sector, systematic literature review, public value, AI governance, leadership, administrative discretion, digital government, foundation models, algorithmic bias, street-level bureaucracy