AI can boost municipal productivity, but only when local governments have the digital maturity and governance mechanisms to manage it; without readiness, algorithmic opacity, data dependence and coordination failures can amplify systemic risks across public services.

Main Finding

AI adoption in local government can raise public sector productivity (as measured via total factor productivity) but those gains are highly conditional on organisational readiness — digital maturity, workforce capabilities, governance quality, and inter-organisational coordination. Without those complementarities, AI increases systemic vulnerability: algorithmic opacity, cybersecurity and data-dependence issues, and coordination failures can generate cascading disruptions across interconnected public-service networks. Resilience — the capacity to anticipate, absorb, adapt, and recover from AI-related disruptions — is therefore a crucial mediating factor between AI adoption and sustained productivity.

Key Points

• Evidence base: systematic review of 68 peer‑reviewed empirical studies (2015–2025).
• Productivity effects are heterogeneous and context dependent; positive effects appear when AI is embedded alongside organisational and institutional complements.
• Organisational readiness dimensions that shape outcomes: digital infrastructure, data quality and stewardship, workforce skills and training, governance arrangements, and cross‑institutional coordination.
• Main risks identified: algorithmic opacity and bias, cybersecurity vulnerabilities, high data dependence, single‑point failures, and coordination frictions that allow disruptions to propagate through networked public systems.
• Resilience framed as four capacities: anticipate, absorb, adapt, recover. Stronger resilience reduces the downside of AI-related shocks and preserves service continuity and public value.
• Conceptual contribution: integrates productivity economics (TFP), systems theory, and public governance into a sociotechnical framework that explicitly links productivity, systemic risk, and resilience.
• Practical output: a conceptual measurement framework and policy guidance aimed at municipal decision makers seeking to balance productivity improvements with risk mitigation and resilience building.

Data & Methods

• Review protocol: PRISMA 2020 methodology for systematic literature reviews.
• Scope: peer‑reviewed empirical studies published from 2015 through 2025 focused on AI in public-sector/local-government settings (or with relevance to networked public administration).
• Outcomes assessed: productivity (often discussed in TFP terms or operational efficiency proxies), methodological quality, reported effect sizes, and contextual/moderating factors relevant to municipal governance.
• Synthesis approach: comparative assessment of empirical findings, methodological quality appraisal across studies, and integration into an integrative conceptual framework. (The study emphasizes transparency and reproducibility through PRISMA procedures; heterogeneity across methods and contexts limited simple meta‑analytic aggregation.)

Implications for AI Economics

• Measurement and modelling
  • Productivity measurement in public sectors should incorporate TFP concepts while accounting for public‑value objectives and nonmarket outputs; standardise metrics and report complementarities (digital capital, skills, governance).
  • Economic analyses should internalise systemic externalities and network spillovers from algorithmic failures, cybersecurity incidents, or data outages — not treat organisations in isolation.
• Policy and investment
  • Investments in AI should be paired with investments in organisational readiness: digital infrastructure, data governance, workforce training, and interoperable institutional arrangements to capture productivity gains.
  • Risk mitigation (transparency, auditability, robust cybersecurity, distributed architectures) is an economic priority because unmanaged tail risks impose negative spillovers across jurisdictions and services.
• Research priorities
  • More causal, longitudinal, and cross‑jurisdictional studies are needed to identify effect sizes and mechanisms, including natural experiments and quasi‑experimental designs.
  • Network and systems‑level empirical work is needed to quantify propagation of shocks and to evaluate resilience interventions (e.g., redundancy, decentralisation, contingency protocols).
  • Incorporate resilience valuation into cost–benefit frameworks to weigh productivity gains against systemic risk reductions.
• Policy design for AI economics
  • Regulatory and governance frameworks should encourage institutional complementarities (e.g., conditional grants for digital maturity), require risk disclosure and stress testing for critical public AI systems, and promote interoperable standards to limit contagion risk.
  • Economic incentives should align with maintenance of public value (equity, continuity, quality) rather than short‑term efficiency gains alone.

Short takeaway: AI can raise municipal productivity, but economics and policy must treat AI as a sociotechnical investment whose returns depend on institutional complements and whose risks are systemic — requiring measurement, governance, and resilience building at the network level.