TLDR: This research paper explores how Artificial Intelligence (AI) can enhance participatory budgeting in smart cities, focusing on Sao Paulo’s ‘Orçamento Cidadão’ initiative. It identifies challenges like declining participation, unfeasible proposals, and weak monitoring. The study proposes AI solutions such as predictive engagement analysis, virtual assistants for proposal drafting, automated feasibility checks, intelligent monitoring, and adaptive digital platforms to overcome these issues. It emphasizes the need for robust state capacities, including open data infrastructure, multidisciplinary teams, and transparent algorithmic governance, to successfully implement AI and foster more inclusive and effective citizen participation.
The integration of Artificial Intelligence (AI) into urban governance is a topic of growing importance, especially in the context of smart cities and citizen participation. A recent study, titled Exploring AI Capabilities in Participatory Budgeting within Smart Cities: The Case of Sao Paulo, delves into how AI can significantly enhance participatory budgeting processes, using the vibrant metropolis of Sao Paulo, Brazil, as a key case study. Authored by Italo Alberto Sousa, Mariana Carvalho da Silva, Jorge Machado, and José Carlos Vaz, the research highlights both the immense potential and the critical requirements for successful AI implementation in public administration.
The Challenge of Participatory Budgeting in Sao Paulo
Sao Paulo’s ‘Orçamento Cidadão’ (Citizen Budget) initiative, which allows residents to propose and vote on how a portion of the municipal budget is spent, faces several hurdles. Despite using an internationally recognized platform (CONSUL, adapted as ‘Participe Mais’), participation has seen a sharp decline, with votes dropping by 86% between 2021 and 2025. This low engagement, representing only 0.05% of the city’s 11 million-plus inhabitants, raises concerns about representativeness. Furthermore, participation is uneven, with central and wealthier districts showing higher engagement, partly due to digital exclusion where mobile access, common in peripheral areas, is underutilized on the platform.
Another significant issue is the high rate of technically unfeasible proposals—around 30% in 2025. Citizens often struggle to formulate proposals that align with budgetary and legal constraints, leading to frustration. The monitoring and accountability of approved projects also present weaknesses; a 2024 report showed that while 89% of projects were marked as ‘completed’ in municipal systems, only 37% were actually implemented on-site. This discrepancy, coupled with a lack of timely updates, erodes public trust.
AI as a Catalyst for Democratic Innovation
The research proposes that AI can address these challenges across five critical dimensions:
1. Expanding Engagement and Reducing Inequalities: AI can analyze participation patterns and socioeconomic indicators to target mobilization campaigns in underrepresented areas. Chatbots integrated with popular messaging apps like WhatsApp could provide personalized notifications and explain processes in accessible language. Recommendation algorithms could suggest relevant content, increasing user interest.
2. Improving Technical Feasibility and Quality of Proposals: Automated pre-analysis systems could cross-reference proposals with urban plans and budgetary laws, offering immediate feedback on incompatibilities. Natural Language Processing (NLP) could translate colloquial ideas into technical administrative language, while machine learning tools could suggest adjustments or viable alternatives.
3. Strengthening Transparency and Monitoring: NLP techniques could automatically extract deadlines and targets from reports, generating alerts for delays. Computer vision systems could compare ‘before and after’ images from georeferenced photos to verify project progress. Predictive analytics could identify chronic delay patterns, enabling proactive interventions.
4. Democratizing Proposal Development: Generative AI systems could act as intelligent assistants, helping citizens structure proposals, suggest missing elements, alert about legal requirements, and even provide cost estimates based on similar past projects. Intelligent databases could flag conflicts with existing urban planning.
5. Modernizing the Digital Platform and Inclusion: Adaptive design systems could optimize the ‘Participe Mais’ interface for mobile devices. Content personalization algorithms could tailor language and information detail to user profiles. AI-based accessibility tools, such as automatic audio transcription or image descriptions, could broaden reach to visually impaired users or those needing foreign language translation.
Also Read:
- AI-Powered Urban Centers Reshaping Daily Life and Governance
- Organizations Redefine Governance to Mitigate AI Risks and Drive Prosocial Innovation
Building State Capacities for AI Implementation
For these AI solutions to be effective, the study emphasizes the need for specific state capacities. These include developing a robust infrastructure of open and interoperable data, integrating various municipal databases (social registries, income indicators, urban laws). Multidisciplinary teams, combining expertise in data science, political science, law, ethics, and communication, are crucial to mediate between technology and citizens. Furthermore, transparent algorithmic governance frameworks are essential to ensure the auditability of AI systems, guaranteeing fairness, explainability, and accountability, and avoiding excessive dependence on private vendors. Continuous investment in digital inclusion and training for public servants is also vital to keep pace with technological evolution and mitigate risks.
In conclusion, the research posits that AI can serve as a powerful tool for democratic equalization in participatory budgeting, transforming processes into more inclusive and deliberative spaces. However, its success hinges on careful planning, robust infrastructure, skilled human capital, and ethical governance, ensuring that technology enhances, rather than replaces, human participation.
