From Sensing to Acting: The Urban Delegation Chain and the Governance of Urban AI

Artificial intelligence in cities is usually judged as a tool by asking whether it improves prediction, coordination and service delivery. This conceptual review argues that such judgements miss a more consequential change: the redistribution of authority to define urban problems, organise knowledge and initiate interventions. Drawing on research on smart cities, algorithmic governance, platform urbanism, urban AI and agentic systems, it treats these as overlapping configurations rather than successive technological stages. In each, governing authority becomes embedded in measurement categories, administrative thresholds, proprietary infrastructures, simulation assumptions and permissions to act. The review develops the concept of the urban delegation chain to explain how small, apparently technical transfers of analytical and operational work accumulate into substantial shifts in public control. It distinguishes representational, decisional, infrastructural, epistemic and operational delegation, and argues that legitimacy depends on the relationship between delegated capability and an institution's effective capacity to challenge and revise it. Technical accuracy, nominal human oversight and transaction logs are each insufficient; what matters is whether authorisation, situated validation, contestability, interruptibility and remedy hold together across the chain. Five propositions translate the framework into a research agenda. Urban AI, the review concludes, should be judged by whether it strengthens a city's capacity to govern collective priorities, including its capacity to refuse, suspend or redesign automation.