Automated state systems boost administrative efficiency but erode democratic legitimacy: AI improves routine tasks and fiscal forecasting yet opaque algorithms (notably Australia’s Robo‑debt) produce procedural injustice. Without value-based transparency and role-sensitive explainability, the rise of algorithmic governance deepens an AI Core–Global South divide and risks long-term digital dependency.

Main Finding

AI as a General-Purpose Technology is driving a structural reconfiguration of state capacity and of how governments manage economies. The reviewed literature (2018–2026), synthesized under PRISMA principles, documents a shift from "street-level" bureaucratic discretion to "system-level" algorithmic architectures — an institutional transfer the review labels "Artificial Discretion." This transfer yields measurable efficiency gains (especially for routinised administration and fiscal forecasting) but generates an "Efficiency–Legitimacy Paradox": automating adjudication and enforcement can improve throughput while simultaneously eroding procedural justice and democratic legitimacy unless transparency and accountability are rebuilt on value-based terms.

Key Points

• Artificial Discretion: governments are reallocating decision authority from frontline public servants to algorithmic systems, changing who (and what) exercises discretion in public policy.
• Efficiency gains: robust evidence of improved processing speed, standardization, and forecasting accuracy in administrative tasks and fiscal management when AI/automated systems are deployed.
• Efficiency–Legitimacy Paradox: bureaucratic efficiency improvements are frequently offset by legitimacy costs — opaque decision logic, errors with systemic effects, and reduced avenues for citizen contestation.
• Black-box harms: opaque algorithmic systems undermine procedural justice; high-profile cases (e.g., Australia’s Robo-debt) exemplify how automated enforcement can violate rights, destroy trust, and provoke costly political backlash.
• Geopolitical divergence: a clear split between "AI Core" states (that develop and control advanced algorithmic governance capabilities) and many countries in the Global South, which face risks of digital dependency, reduced digital sovereignty, and asymmetric leverage in economic governance.
• Visibility mechanisms: public algorithm registers, role-sensitive explainability, mandated audits, and institutionalized human oversight are recurring policy prescriptions shown to mitigate legitimacy harms and restore citizen trust.
• Governance transition: sustainability of an "algorithmic state" depends less on pure technical fixes and more on institutional reforms — shifting from technocratic secrecy toward value-based transparency, accountability, and algorithmic justice.

Data & Methods

• Review scope: systematic review of peer-reviewed studies, policy reports, and empirical cases published 2018–2026 addressing AI-driven governance, policy integration, economic consequences, and democratic legitimacy.
• PRISMA-based workflow: the synthesis followed structured search, screening, eligibility, and inclusion stages to identify relevant empirical and theoretical work across disciplines (public administration, political science, economics, law, and STS).
• Evidence types: mixed evidence base including qualitative case studies (e.g., Robo-debt), comparative cross-country analyses, quantitative evaluations of administrative efficiency and fiscal forecasting, survey-based trust measures, and normative/legal analyses.
• Synthesis methods: thematic synthesis for qualitative findings; where comparable quantitative outcomes existed (e.g., processing time, forecasting error), aggregated comparisons and meta-analytic summaries were used to assess direction and robustness of effects.
• Limitations noted in source literature: heterogeneity in outcome measures, under-reporting of negative results, selection bias toward high-profile cases, and uneven geographic coverage (concentration on AI Core countries).

Implications for AI Economics

• Public finance and fiscal capacity: AI can improve tax administration, benefits delivery, and fiscal forecasting, potentially increasing revenue mobilization and reducing leakage — but legitimacy losses and subsequent protests/litigation can negate fiscal gains or raise enforcement costs.
• Labour and skills in the public sector: automation will reallocate tasks away from routine public-facing roles toward oversight, audit, and policy design roles — creating transitional displacement risks and requiring reskilling and institutional redesign.
• Distributional outcomes: automated governance can entrench biases and create asymmetric harms (e.g., false positives in enforcement) that disproportionately affect vulnerable populations, increasing inequality and welfare losses unless countervailing measures are implemented.
• Transaction and political costs: while algorithms can lower administrative transaction costs, legitimacy deficits can increase political transaction costs (appeals, litigation, policy reversals), affecting the net economic benefit of automation.
• Sovereignty and geopolitical economy: dependence on external AI platforms and proprietary tools concentrates technological rents in AI Core economies, risks digital dependency for the Global South, and can exacerbate unequal bargaining power in trade, aid, and infrastructure finance.
• Market structure and rents: concentration of algorithmic governance platforms could create new sources of monopoly or oligopoly rents with implications for public procurement, competition policy, and the fiscal burden of purchasing/adopting governance AI.
• Policy prescriptions with economic relevance:
  • Invest in domestic AI capacity and intergovernmental cooperation to mitigate digital dependency and preserve policy sovereignty.
  • Treat algorithmic transparency (registers, standardized explainability) as a public good that reduces legitimacy risks and potentially lowers long-run enforcement costs.
  • Build institutional accountability (audits, impact assessments, role-sensitive explainability) into procurement and deployment to internalize social costs and protect fiscal returns from automation.
  • Fund reskilling programs and social safety nets to manage labor transitions within the public sector and maintain service quality.
  • Incorporate political and legitimacy risk assessments into cost–benefit analyses for AI adoption in government, not just technical efficiency metrics.

Overall, the economic promise of AI for governments is substantial, but realization depends on governance choices: preserving legitimacy, distributing gains fairly, and building sovereign capacity will determine whether AI strengthens state capacity sustainably or produces short-term efficiency that yields long-term political and economic liabilities.