What If AI Systems Self-Operate Our Cities?

Imagine a future where AI systems manage our cities. This text delves into this idea of a “synthetic city” by introducing the concept of self-operating governance. Before providing its meaning, we will explore its ontogeny by analyzing how urban governance has evolved over the development of digital technologies. Over the past few decades, the rise of digital technologies and the concept of the “smart city” have profoundly transformed how cities are managed.¹Lorena Melgaço and Katharine S. Willis, “Editorial: Social Smart Cities: Reflecting on the Implications of ICTs in Urban Space,” plaNext – next Generation Planning 4 (2017): 5–7, https://doi.org/10.24306/plnxt.2017.04.001. This evolution can be seen in the emergence of “smart” urban governance, which reflects the gradual integration of technologies into urban systems. It has led to the emergence of several analytical frameworks, including digital governance, smart governance, and algorithmic governance.

At its core, digital governance involves the use of information and communication technologies (ICTs) to digitize administrative processes, improve the efficiency of public services and facilitate citizens’ access via digital platforms.²Erico Przeybilovicz and Maria Alexandra Cunha, “Governing in the Digital Age: The Emergence of Dynamic Smart Urban Governance Modes,” Government Information Quarterly 41, no. 1 (2024): 101907, https://doi.org/10.1016/j.giq.2023.101907. For example, in 2009, Amsterdam implemented a platform that allowed municipal councils to manage infrastructure such as smart lighting. This lighting reduces energy consumption based on real-time pedestrian data and enables collaborative urban planning. Smart governance deepens this dynamic by using more advanced technologies. These include the Internet of Things (IoT) and big data, allowing the exploitation of data in real time, anticipating needs, and guiding public action.³Przeybilovicz and Cunha, “Governing in the Digital Age.” For example, under the Smart City Barcelona⁴“Smart City Barcelona: a network of networks”, SGD Local Action, accessed August 24, 2026, https://sdglocalaction.org/smart-city-barcelona/. since 2012, the city has deployed 19,500 smart meters to optimize energy use. Sensors on waste bins have improved waste collection routes (see Figures 1 and 2). They have also improved public transportation with real-time bus arrival information and smart parking guidance.

These examples illustrate that governance then becomes more adaptive, predictive, and participatory, transforming the city into a real “platform of platforms.” However, in these two approaches, technology primarily serves as a tool, supporting human decision-making without fundamentally altering its structure.

This gradual evolution has paved the way for algorithmic governance (illustrated in Figure 3). With the rise of AI, increased computational capabilities and massive data collection, technologies are no longer just supporting public action, but are beginning to structure decisions themselves. Algorithmic governance, or algocracy, describes the integration of AI systems into urban decision-making processes.⁵Richard Weiskopf and Hans Kause Hansen, “Algorithmic Governmentality and the Space of Ethics: Examples from ‘People Analytics,’” Human Relations 76, no. 3 (2023): 483–506, https://doi.org/10.1177/00187267221075346. It is based on urban AI systems defined as systems integrating data from the urban environment, processed by algorithms, and whose results find useful applications in the socio-spatial fabric of the city.⁶Tim Heinrich Son, Zack Weedon, Tan Yigitcanlar, Thomas Sanchez, Juan M. Corchado, and Rashid Mehmood. “Algorithmic Urban Planning for Smart and Sustainable Development: Systematic Review of the Literature,” Sustainable Cities and Society 94 (2023): 104562, https://doi.org/10.1016/j.scs.2023.104562. For example, traffic management systems can adjust traffic lights in real time. Predictive tools can anticipate infrastructure breakdowns or crime risks.

In this regime, AI systems now influence the form, rhythm, and internal logic of public action. Some authors call this algorithmic governmentality.⁷Weiskopf and Hansen, “Algorithmic Governmentality and the Space of Ethics.” Decision-making is increasingly influenced by technological systems that incorporate mathematical models, embedding urban governance in a “society of numbers” where the city can be quantified.⁸Michael Batty, The Computable City: Histories, Technologies, Stories, Predictions (The MIT Press, 2024). This transformation also promotes the emergence of anticipatory governance based on the ability of AI systems to predict and preempt urban phenomena.⁹Federico Cugurullo and Ying Xu, “When AIs Become Oracles: Generative Artificial Intelligence, Anticipatory Urban Governance, and the Future of Cities,” Policy and Society 44, no. 1 (2025): 98–115, https://doi.org/10.1093/polsoc/puae025.

Thus, the evolution of urban governance reveals a gradual shift from a logic of digitization of services to a logic of anticipation and autonomy in decision making.¹⁰Federico Cugurullo, “Urban Artificial Intelligence: From Automation to Autonomy in the Smart City,” Frontiers in Sustainable Cities 2 (2020), https://doi.org/10.3389/frsc.2020.00038; Cugurullo and Xu, “When AIs Become Oracles.” This path raises the question of how far this technological integration can go. In other words, what would a city look like if its management were predominantly, if not entirely, controlled by AI systems?

Self-operating systems leading the future of urban management

To answer this question, this paper proposes the concept of self-operating governance (illustrated in Figure 4). In the following sections, we will answer this question with two things. First, we will present the contours of this concept and the signals that draw it out. Secondly, we will illustrate this form of governance in a prospective scenario concerning fire management in municipalities. This will enable us to obtain the views of city managers on its feasibility, opportunities, and challenges.

The notion of self-operating governance fits within the Synthetic City paradigm and emphasizes the growing capacity of algorithmic infrastructures to execute, coordinate, and adjust certain urban operations. It does not refer to complete autonomy of AI systems, but rather to functional autonomy in the execution of certain operations related to urban services delivery. It describes a form of urban governance in which AI-driven prediction and autonomous management systems interact and evolve, leading to a city managed with minimal human intervention. AI systems drive the provision of services based on the metrics they create, analyze, and optimize themselves. Urban systems now run based on the internal consistency of a computational system that can adapt its actions based on its own feedback. This structure holds great potential, as evidenced by studies on the advantages of AI in urban settings.¹¹Federico Caprotti et al., “Why Does Urban Artificial Intelligence (AI) Matter for Urban Studies? Developing Research Directions in Urban AI Research,” Urban Geography 45, no. 5 (2024): 883–94, https://doi.org/10.1080/02723638.2024.2329401; Walter Leal Filho et al., “The Role of Artificial Intelligence in the Implementation of the UN Sustainable Development Goal 11: Fostering Sustainable Cities and Communities,” Cities 150 (2024): 105021, https://doi.org/10.1016/j.cities.2024.105021. These advantages include enhanced responsiveness, constant optimization, uniform and swift handling of demands, and rapid service delivery. But the same algorithmic rationalization can transform administrative mediation into a standardized flow. Access to services is limited based on the citizen’s aptitude to input data into the system’s required format.

This is an emerging regime in which public decision-making is no longer only assisted by technical devices. Indeed, it is co-produced by infrastructure capable of anticipating, operating, and feedback on their own operating conditions. Self-operating governance is thus characterized by a gradual reduction of human centrality in the daily management of cities, in favor of integrated AI systems. Humans intervene less as decision-making agents than as supervisors of a set of dynamics already structured by algorithmic infrastructures. It no longer describes AI-assisted governance, but rather co-produced, co-optimized and co-administered governance.

It is important to stress, however, that self-operating governance is not a stabilized reality. It is an analytical horizon in the synthetic city paradigm. Furthermore, it is a prospective theory that allows us to grasp the direction that cities are taking in the face of the increasing integration of AI systems into their operations. This aligns with the rapid pace of progress and algorithmic performance, as well as the increasing interconnection of generative and predictive models. This notion provides a framework for interpreting current trends without presupposing their future outcome or consolidation.

This model is based on a close articulation between two complementary capabilities of AI systems. On the one hand, the predictive capabilities of machine learning and deep learning (ML/DL) systems, and on the other hand, the operational and generative capacities of large language models (LLMs). This interplay results in a feedback loop where anticipation and optimization mutually reinforce each other. On the other hand, automated operations continuously generate structured data. Every citizen interaction and urban activity (request, traffic, complaint, permit application, etc.), every response provided, every file processed or classified becomes a usable trace. This data is used as input into predictive models that can identify trends, such as an increase in complaints related to a specific infrastructure or service. This allows for operational optimization to serve as a source of learning for anticipation. Conversely, predictive models directly influence automated operations. For instance, predictive maintenance systems can foresee failures in water or transportation systems. These forecasts are subsequently incorporated into operational systems. LLMs adjust treatment priorities, autonomously formulate intervention instructions, or customize responses to citizens according to predicted hazards. In this context, LLMs no longer just produce answers. Instead, they reconfigure their processing chains in response to predictions from forecasting systems. This involves prioritizing time-sensitive requests, tailoring language to user profiles, or even automating certain routine administrative tasks. This dynamic gives rise to a self-referential computational regime. It is a feedback loop in which AI systems generate the data that drives their own functioning. Gradually, they shape the direction, importance, and methods of urban governance. This loop doesn’t eradicate humans, but instead sidelining their role in the decision-making process, positioning them as overseers rather than producers.

Self-operating governance is taking shape through a series of converging weak signals revealing a gradual transformation of the urban management model.

• The first factor is the gradual tokenization of urban governance. Algorithmic systems are increasingly used to deliver municipal services¹²Cugurullo and Xu, “When AIs Become Oracles.”. Each citizen’s request, whether it is a complaint, a request for service, or a request for information, becomes a standardized computational unit integrated into an algorithmic chain. Language models and automation systems modulate, classify, and guide municipal action.¹³Yu Zheng et al., “Urban Planning in the Era of Large Language Models,” Nature Computational Science 5, no. 9 (2025): 727–36, https://doi.org/10.1038/s43588-025-00846-1. Each citizen interaction, from the initial request to the provision of the service, thus becomes an algorithmic cycle that produces new data. In this logic of permanent optimization, this data feeds the training of algorithmic models that can better predict, anticipate, and accelerate the processing of requests.
• The second factor is the emergence of an epistemic ecosystem of AI in urban management and its growing institutional anchoring. AI systems are developing within a network of multilevel actors made up of municipalities, technology companies, non-profit organizations, research centers, and public agencies.¹⁴Ida Farida et al., “Responsible Urban Innovation Working with Local Authorities a Framework for Artificial Intelligence (AI),” Scientific Journal of Informatics 10, no. 2 (2023): 121–26, https://doi.org/10.15294/sji.v10i2.43818. Collectively, these entities create tools, standards, and technological visions that gradually establish the role of AI in urban management. Municipal managers generally have a favorable view of these technologies and are open to their integration into municipal operations.¹⁵George Burciaga, “AI as an Agent of Change: Reimagining Cities Through Mayor-Led Innovation,” City, July 23, 2025, https://www.tomorrow.city/ai-agent-of-change-american-cities/. This vision, which is both technoptimistic and pragmatic, helps to legitimize a gradual integration of AI into urban management. Several structural factors reinforce this dynamic, including the search for organizational performance, mimetic pressures between municipalities, and the growing idea that AI is not only a useful but inevitable technology.¹⁶A. Paula Rodriguez Müller, Luca Tangi and Amandine Lerusse, “Understanding the Adoption of Artificial Intelligence in Local Government Decision‐Making: The Influence of Institutional Pressures and Managerial Perceptions,” Public Administration (2025): 1-17, https://doi.org/10.1111/padm.70033; Sajani Senadheera et al., “Understanding Chatbot Adoption in Local Governments: A Review and Framework,” Journal of Urban Technology 32, no. 3 (2025): 35–69, https://doi.org/10.1080/10630732.2023.2297665; Tan Yigitcanlar, Anne David, Wenda Li, Clinton Fookes, Simon Elias Bibri and Xinyue Ye. “Unlocking Artificial Intelligence Adoption in Local Governments: Best Practice Lessons from Real-World Implementations,” Smart Cities 7, no. 4 (2024): 1576–625, https://doi.org/10.3390/smartcities7040064. In a context of international competition for the economic benefits of AI, cities are also becoming laboratories for technological experimentation.¹⁷Rachel Adams, The New Empire of AI: The Future of Global Inequality (polity, 2025); “AI-powered cities of the future”, Deloitte, accessed August 24, 2026, https://www.deloitte.com/global/en/Industries/government-public/research/ai-powered-cities-of-future.html. This accelerates the integration of these systems. The potential savings in time, resources and efficiency are strong incentives for municipal governments.¹⁸Rodriguez Müller et al., “Understanding the Adoption of Artificial Intelligence in Local Government Decision‐Making.” When a technology ceases to be perceived as an innovation and becomes the environment in which institutions operate, it silently transforms organizational logics.¹⁹Cugurullo and Xu, “When AIs Become Oracles.” AI systems are becoming naturalized, stabilized, and routinized in urban management. They become the conditions for the possibility of governance itself. As they become deeply integrated into operations, it becomes difficult, if not impossible, to remove them. The city is now adjusting to the constraints that they impose. This is where the governance of algorithms comes in. It is not about abandoning technology, but rather about harnessing it in a more responsible manner.
• The third factor is the rapid evolution of AI architectures. The accelerated development of specialized models, combined with the emergence of new architectures such as world models, is a major transformation. These systems are able to dynamically represent their environment and anticipate the consequences of different actions²⁰Yutaka Matsuo et al., “Deep Learning, Reinforcement Learning, and World Models,” Neural Networks 152, no. C (2022): 267–75, https://doi.org/10.1016/j.neunet.2022.03.037.. As some AI researchers such as Yann LeCun point out, these architectures lay the foundations for AI capable of perceiving, anticipating, planning, and acting autonomously. Transposed to the urban context, these technologies offer the computational foundations of a system that can continuously learn from the data produced by urban infrastructure, adjusting its strategies and correcting its decisions without constant supervision. Although these capabilities are currently limited in practice, they suggest a potential path towards the development of integrated systems that can combine perception, forecasting and action into a unified operational structure.

What do city managers think about it?

Now, let’s illustrate this form of governance and explore the opinions of city managers on it. To do this, we created a hypothetical scenario (see Figure 5), titled “When AI systems take over fire management”, describing a fire management crisis in municipalities. This scenario involves collaborative action by various AI systems in a self-operating logic. We presented the scenario to 11 city managers from Quebec municipalities in Canada. Then we used a reflexive approach to collect their perceptions regarding 3 themes: feasibility (is the scenario realistic in the time frame?), opportunities (do you see any benefits from it, and why?), and weaknesses/threats (what are the challenges or the pitfalls?).

The scenario: When AI systems take over fire management

It’s March 10, 2040 at 3:23 p.m. when a thermal sensor in a residential building in the city center detected an abnormal rise in temperature. The alert is automatically transmitted to the central urban intelligence system. In a few milliseconds, the AI confirms the fire by cross-referencing images from surveillance cameras and smoke sensors.

Immediately, the AI deploys a convoy of autonomous fire trucks, accompanied by drones and fire robots. Vehicles are given an optimized trajectory, while traffic lights automatically adjust to give them priority.

On site, the drones map the building in real time, identify critical areas and start extinguishing the flames using fireproof foam. Robots intervene inside to contain the fire and secure exits, while human firefighters supervise and intervene only in delicate rescue situations.

At 3:47 p.m., the fire has been brought under control. The central system automatically generates a detailed report for municipal services: response time, resources mobilized, material damage, condition of victims. In the next minute, an official statement from the mayor, written by the virtual communication assistant, is published on the municipal website and relayed on social networks.

At the same time, citizens receive an alert on their municipal app, informing them that the situation is under control and specifying the areas to avoid.

City managers’ perception

City managers agree that the concept is technologically feasible in light of rapid technological development. One noted, “it’s really not impossible… it’s realistic.” Other city managers have highlighted that some of the technologies described in the scenario already exist, and some municipalities are even deploying some of them. “…drones or autonomous robots already exist. They are still in their early stage but they are already available… Synchronized traffic lights I think are even under deployment in some Quebec municipalities, like trois rivieres or Quebec City.” Self-operating systems are characterized by the interconnectedness of various subsystems to automate municipal operations. All city managers agree that automation of processes, including their physical dimensions (such as robotics and autonomous vehicles), is an almost inevitable evolution of contemporary urban systems. However, there is no consensus on the timeline for this transformation. Several managers believe that “2040 is too early.” They cite constraints related to technology readiness, organizational capacity, and regulatory frameworks. However, some of them also believe that this time frame is realistic, given the rapid progress of AI technologies. As one of them stated, “…15 years when it comes to innovation is an eternity.”

In terms of opportunities, all the managers shared a common vision of AI systems as a lever for optimizing municipal services and managing urban growth. One highlighted that “it is the employee you don’t need to hire.” However, the interviews also revealed that the acceptance of self-operating systems hinges primarily on their ability to improve operational efficiency. Participants highlight the potential benefits of optimizing processes, reducing intervention times, and streamlining decision-making chains. As one participant pointed out, “it would be a great gain for municipalities to optimize these processes and automate more.” This perspective suggests that AI systems are increasingly viewed as infrastructure capable of orchestrating various administrative and informational operations. A participant noted, “In the long term, we have a strategic vision… there will be more business processes that will be automatable with AI… an agent will be able to call tools in the organization to fetch information and perform actions in systems.” From this perspective, self-operating governance is perceived as an extension of the current dynamics of integration and interconnection of municipal services. There is a strong focus on automating information exchange, streamlining interdepartmental collaboration, and increasing accountability. This dynamic is driven by the growing demands on municipal governments, while resources remain limited. One participant explained, “I have an increase of nearly 400% in citizen requests, but I haven’t had more employees for five years.” In this context, AI is seen as a way to cope with this pressure by automating certain tasks and improving team productivity and time gains. As a participant pointed out, “There is a great waste of time at the moment in the municipal field…. To be forced to spend time behind a computer, to calculate deals, when we are able to optimize and automate certain processes.” This aspiration is also reflected in the desire to use AI to transform urban management practices. One participant took the example of citizen service delivery and highlighted the need for more automation. “What we would also like in the future is for citizens to get into the habit of using our chatbot before coming to talk to us or come to see us. There is a lot of time that is taken up to respond to citizens’ requests when they could very well go and get the information on their own. The chatbot has the answer to that.”

Regarding weaknesses and threats, despite adhering to the principles of optimization, city managers strongly emphasized the importance of human involvement in urban governance. Many of them expressed reservations about the full automation of services, particularly because of the issues of trust, responsibility and relationship with citizens. As one participant pointed out, “as a society it will lack a little human warmth.” This highlights an important limitation of self-operating systems. Indeed, while automation is desired for technical and repetitive processes, it is perceived as problematic when it affects human interactions or critical situations requiring judgment and empathy. Moreover, the interviews highlight a keen awareness of the risks associated with these systems, particularly in terms of model alignment, error management, and data governance. City managers recognize that AI systems are inherently flawed and require oversight mechanisms. The idea of a fully autonomous AI system is thus largely rejected in favor of a model in which humans retain a role of validation, control, and arbitration. This approach aligns with a “human-in-the-loop” governance philosophy, whereby automation is regulated and managed by institutional and organizational frameworks. In this logic, city managers favor an incremental approach to AI integration based on experimental, testing, and adjustment phases. Current applications are primarily confined to administrative or low-risk duties, with high-risk areas like public safety receiving more care and scrutiny. This progressive strategy reflects the desire to control risks and seize the opportunities that AI offers.

Overall, these results suggest that self-operating urban governance is not yet a stabilized empirical reality. It is more of an emerging trajectory structured by optimization dynamics, institutional constraints, and socio-technical trade-offs. It thus appears to be a transformation horizon to which municipalities are tending, without calling into question, at this stage, the centrality of the human element in public decision-making.

Implication for urban policies

Self-operating systems alter the political, social, ethical, and economic aspects of city management, prompting a thorough examination of the current paths of AI integration in urban settings.

The initial matter to address is the fundamental question of who controls the city. If urban operations become self-operated by AI systems, who really governs? This query demonstrates that self-operating governance cannot be considered as an absolute regime. In reality, a city entirely driven by AI systems would be a utopian vision. Urban governance remains intrinsically political because it defines orientations, priorities, and arbitrations in a multi-stakeholder architecture. In this context, self-operating governance should be understood as a specific operational management modality, in which certain functions are delegated to AI systems, under the authority of public actors.

However, this delegation raises questions about real power, especially since the technologies used are mostly designed by large technology companies. Research on “firm-states,”²¹ Juho Lindman, Jukka Makinen and Eero Kasanen, “Big Tech’s Power, Political Corporate Social Responsibility and Regulation,” Journal of Information Technology 38, no. 2 (2023): 144–59, https://doi.org/10.1177/02683962221113596. and the corporate nature of smart cities²²Rob Kitchin, Claudio Coletta, Leighton Evans, Liam Heaphy, and Darach Mac Donncha. “Smart Cities, Algorithmic Technocracy and New Urban Technocrats,” in Planning and Knowledge, ed. Mike Raco and Federico Savini (Policy Press, 2019), https://doi.org/10.1332/policypress/9781447345244.003.0015. shows that these companies are exerting an increasing influence on public policies, including at the urban level. This raises concerns about the autonomy of local governments, which are becoming increasingly dependent on technology. Although it can enhance service delivery, it also has the potential to mask a complex web of influence where decisions are subtly shaped by hidden private infrastructures. Policy-makers should prioritize public oversight to prevent vendor lock-in and ensure democratic control.

Self-operating governance is part of a technophile vision of the urban space as a synthetic city.²³Javier Echeverría and Raúl Tabarés, “Artificial Intelligence, Cybercities and Technosocieties,” Minds and Machines 27, no. 3 (2017): 473–93, https://doi.org/10.1007/s11023-016-9412-3. However, public perception of AI is complex and often uncertain, with some citizens advocating for a more cautious approach due to potential risks. Issues such as surveillance, datafication, digital rights protection, and environmental impact become particularly crucial in this model. Moreover, the reduced need for human intervention in administrative and operational processes raises many concerns related to employment and work transformation. Self-operating governance cannot, therefore, be reduced to a logic of optimizing services because it raises a fundamental question for society. It requires a trade-off between operational performance and collective values, involving democratic debates on the type of city and society desired. In this regard, urban policies must protect digital rights, address labor displacement, and redistribute technological efficiency gains to ensure equitable urban development.

Finally, this regime also raises central ethical issues related to the place of humans in the city. If decisions and actions are left to AI algorithms, the question arises: who is truly in charge? Is the city still a human-centric environment, or is it evolving into a technological network where people are relegated to data points? In this scenario, human connections are gradually transformed into statistical data feeding into algorithmic loops. This raises questions about agency, responsibility, and human dignity. Unsupervised self-operating governance could thus lead to a form of depoliticization and dehumanization of public action. Therefore, policy-makers must ensure that people remain at the core of decision-making and that mechanisms of control, transparency and accountability are in place.

Thus, self-operating governance and the synthetic city as a whole cannot be understood solely as a technical evolution of urban management. It is a complex issue that alters power dynamics, societal models, ethical standards, and economic systems. As such, its implementation requires a thoughtful and critical approach that can blend innovation with democratic principles.

Leandry Jieutsa

Leandry Jieutsa is a researcher at the UNESCO Chair in Urban Landscape at the University of Montreal. He studies the orchestration of innovation ecosystems to promote responsible AI deployment in local governments. He also works as a consultant in innovation and digital transformation in public administration. Urban photographer and graphic designer, he founded the Africa Innovation Network, a think and do tank helping cities and governments navigate technological and institutional change.