Unlocking AI Adoption in Local Governments: Best Practice Lessons from Smart Cities

1. Introduction

Local government, as the closest level of governance body to the community, occupies a critical position in ensuring efficient and effective service provision (Vincent, 2015; Yigitcanlar et al., 2021a). The use of Artificial Intelligence (AI) technologies in various spheres of local government service delivery has expanded significantly in recent years (Mikhaylov et al., 2018). This adoption spans a broad spectrum of services, from disseminating information to the public, gathering community feedback, and managing complaints, to tax collection, transportation management, water and sewage management, waste collection and management, and the maintenance of the public amenities (Hodgkinson et al., 2017; David et al., 2023; Ye et al., 2023a). As local governments increasingly adopt AI technologies (Engin & Treleaven, 2019), thanks to the local smart city agendas (Yiigtcanlar et al., 2022), understanding their functions and impacts becomes imperative (Buchelt et al., 2024).

Understanding the nuances of AI technology is a crucial process (Criado & Gil-Garcia, 2019; Regona et al., 2024). Different AI technologies are designed to address specific tasks or challenges (Benbya et al., 2020; Jan et al., 2023). The discernment of which technology is best suited for a particular purpose enables organisations to streamline their workflows, processes, and systems, thereby enhancing efficiency and productivity (Bandari, 2019; Lins et al., 2021). As AI technologies progressively evolve and permeate across diverse domains, this understanding becomes increasingly critical for local governments navigating complex technological landscapes to realise their aims (Madan & Ashok, 2022; Habbal et al., 2024). Specifically, delving into the potential of AI technologies to fulfil service delivery objectives and enhance public welfare is of paramount importance.

Despite the rapid growth in AI adoption in smart cities and its potential for positive impact, the scholarly literature offers limited insights into the utilisation of AI technologies in local governments. Moreover, to the best of our knowledge, there exists no studies that comprehensively investigate the practical implementations of AI technologies within local government settings. As such, this study analyses 262 cases in local governments where AI technologies are utilised, employing a grey literature review technique. These instances are chosen to foster a deep comprehension of the phenomenon of local government AI adoption (Farghaly, 2018; Mohajan, 2018), and a grey literature review helps mitigate the publication bias, fostering a balanced picture of available evidence (Polanin et al., 2016; Paez, 2017). Through this approach, the study aims to develop a consolidated understanding of AI technology utilisation in smart city local government service provision, and to generate lessons and best practice for similar smart city initiatives.

2. Literature Background

AI is becoming increasingly pervasive and evolving into an umbrella term that encompasses various technological facets (Lu, 2019; Dwivedi et al., 2021; Gill et al., 2022). This technology set covers a variety of specific segmentations such as computer vision, natural language processing, machine learning (Dong et al., 2021; Zhang & Tao, 2021; Du et al., 2023; Sharma et al., 2022), deep learning, and their generative subsets (Ye et al., 2022; Bibri et al. 2024), extending its influence across a broad spectrum of domains. This broad scope has rendered the term ‘AI’ inherently vague, making its definition elusive (Clarke, 2019). As AI pervades diverse applied urban domains (Bibri et al., 2023a) and industrial spheres (Yigitcanlar et al., 2021b), it becomes imperative to systematically classify AI technologies (Barredo Arrieta et al., 2020; Sarker, 2022) to effectively understand their functionalities, capabilities, and implications.

Technically, AI was first introduced to the academic literature in 1943 (McCulloch & Pitts, 1943); since then, scientists, researchers and philosophers have endeavoured to conceptualise and map out the AI technologies. Notable contributions include those by Anyoha (2017) and Roser (2022), who provide insights into the AI paradigm. Anyoha (2017) delineated the history of AI over seven decades from 1930, while Roser (2022) mapped the AI history of the last eight decades and offered future predictions based on Sevilla et al. (2022).

In the meantime, the AI knowledge map, developed by Corea (2018) and derived by Nesta (2020), provides an overarching understanding of AI, its subtypes, the problem domains in which it can be applied and the capabilities of AI technologies. Figure 1 and Table 1 elucidates the positioning and definition of each technology in the broader AI knowledge realm.

The adoption of AI technologies within the governmental sector is rapidly revolutionising administrative processes, service delivery, and policymaking (Criado & Gil-Garcia, 2019; Engin & Treleaven, 2019). By leveraging AI technologies, government agencies worldwide are streamlining operations, enhancing citizen services, and making data-driven planning and decisions (Dwivedi et al., 2021; Gracias et al., 2023; Li et al., 2023). For example, AI-driven chatbots and virtual assistants are increasingly being used to enhance citizen engagement by offering continuous assistance and tailored support, which can help improve accessibility and responsiveness (Chaturvedi et al., 2023; Kamalov et al., 2023). Furthermore, AI technologies address a range of challenges encountered by government entities, including optimising resource allocation, handling large and diverse datasets, mitigating shortages of experts, managing predictable scenarios, and addressing procedural inefficiencies (Gruetzemacher et al., 2021; Gill et al., 2022). Concurrently, AI is being utilised for citizen inquires and information exchange, including answering questions, assisting with document completion and search, sentiment analysis, routing requests, translation services, and drafting documents (Mehr, 2017; Lovell et al., 2023; Ye et al., 2023b).

Accordingly, the exploration of AI adoption and deployment within the public sector has emerged as an increasingly important area of academic interest. Investigations in this area are wide-ranging, including efforts to understand the different ways in which different public sectors are adopting AI, as well as in-depth studies of the opportunities and challenges presented by AI deployment (Mikhaylov et al., 2018; Desouza, 2019). Scholars also examine the proliferation and trends of AI applications across different public sector domains (Sousa et al., 2019; Maragno et al., 2021), factors influencing AI adoption, and employees’ perceptions of its implementation. Furthermore, research has explored the synergistic benefits of collaboration between the public and private sectors in AI use (Wang et al., 2021), ethical considerations of AI (Wirtz et al., 2018) and more.

Simultaneously, governments operate at multiple levels and each level of government has its own policy priorities, governance structures, and resource constraints (Mwisongo & Nabyonga-Orem, 2016; da Cruz et al., 2018; Sousa et al., 2019; Meuleman, 2021). Academia is interested to understanding how these disparities influence the adoption and implementation processes of AI technologies. Local government, in particular, stands out in this regard. Despite their constrained authority and delineated responsibilities, local governments are increasingly integrating AI into service provision (Yigitcanlar et al., 2021a). Nevertheless, research in this area remains nascent and circumscribed.

Recently, Yigitcanlar et al. (2021a) formulated a conceptual framework for responsible AI in the local government. Moreover, their subsequent studies have expanded their research to capture the city manager’s perceptions in Australia and the US regarding the adoption of AI in local government contexts (Yigitcanlar et al., 2023a), as well as the public perception in Australia and Hong Kong concerning this topic (Yigitcanlar et al., 2023b). Similarly, Distor et al. (2021) explored the attitudes and perceptions of officials within the local governments of the Philippines, while Vogl (2021) analysed the challenges of adopting AI within local government services. Consequently, the findings from these studies indicate the need for further investigation into how AI technologies are being implemented and utilised in local government settings—particularly in the context of smart cities.

3. Research Design

Due to a significant knowledge gap, this study was undertaken to address the research question of ‘how AI technologies are being utilised within local government settings’. Data collection occurred in three distinct stages: (a) Defining the criteria; (b) Searching documents; (c) Filtering documents (Figure 2). The search conducted in January 2024. To derive the findings, a grey literature review methodology was employed.

3.1. Criteria Identification

First, the Boolean search string and inclusion and exclusion criteria were defined. The Boolean string was developed based on the AI knowledge map developed by Corera (2018) (Figure 1) and the study area key words. Accordingly, the strings are developed separately for each AI technology. For example, (“inductive logic programming”) AND ("local government" OR "municipalit*" OR "city*" OR "town" OR "council" OR “borough” OR “Shire”). The used AI technological terms are, (“inductive logic programming,” “robotic process automation,” “expert system,” “fuzzy system,” “decision networks,” “Bayesian program synthesis,” “probabilistic programming,” “neural networks,” “deep learning,” “machine learning,” “ generative adversarial network,” “computer vision,” “natural language processing,” “autonomous system,” “distributed artificial intelligence,” “ decentralised artificial intelligence,” “ affective computing,” “ambient computing,” “evolutionary algorithms,” “genetic algorithms”).

Next, as shown in Table 2, the inclusion and exclusion criteria were established. The information is gathered from websites and scholarly journals, book chapters, and conference proceedings have not been taken into consideration to adhere to the research aim. It is crucial that the cases concentrate on local government service delivery. This study does not consider other public sector agencies and private organisations that operate within the local government area. The timeline was left open in the search criteria for the year of publication, such that the investigation can better understand the evolution in local government AI use and adoption.

3.2. Document Search

The principal data collection platform is the Google search engine. We employed the above-mentioned search query to search each AI technology separately. Document searches produced thousands of hits in several cases. In these situations, the search was carried out up until the Google alert stated: “In order to show you the most relevant results, we have omitted some entries very similar to the 120 already displayed. If you like, you can repeat the search with the omitted results included”. A useful website employed during the Document Search stage was “Govlaunch” (https://govlaunch.com). This is a free wiki for innovative local government. Up to January 2024, the website had 8,922 instances from all around the world. From this website, 129 local government AI use-cases were identified to be included in the analysis.

3.3. Document Filtering

Each case has been systematically recorded within an Excel spreadsheet. For each case, we recorded information against seven categories: (a) AI technology; (b) Local government name; (c) Country name; (d) Comprehensive use description; (e) Year of introduction; (f) URL link to the published webpage; (g) Description of services. Subsequently, an additional Google search was conducted to identify the service offered by local governments in the concerned countries. Accordingly, five main service and 28 sub-service categories have been identified (Table 3).

It was found, to our surprise, that none of the local government AI use-cases from China had been documented at the ‘country name’ filtering phase; despite the top two countries in AI development and adoption being the US and China (AlShebli et al., 2022; Hine & Florida, 2022). Thinking the reason of Chinese local governments only sharing their information on AI utilisation in Chinese language and domestic platforms, we conducted a supplementary search task for cases in China via the Baidu search engine (Chinese search engine) to ensure the comprehensiveness of our database. To reduce the difficulty of the search task, for the case of China, we only searched for information on the local government's official website, i.e., information officially announced on the local government website—‘gov.cn’. In addition, the snowballing strategy was adopted to achieve comprehensive coverage of the cases (Wohlin et al., 2022). Finally, a total of 62 cases in China were identified. Lastly, duplicate entries were removed. This leaved us with 262 AI use-cases from 170 local governments—forming a local government AI use-case dataset (see Appendix A).

4. Analysis and Results

4.1. General Observations

The local government AI use-case dataset developed through the grey literature review in documented in January 2024 encompasses 262 cases, spanning 170 local governments. Despite the existence of numerous AI technologies (Figure 1), the records indicate that only eight technologies are actively utilised within local government settings. Among these, NLP emerges as the most prevalent, accounting for 108 cases, followed by RPA with 58 cases, NN with 47 cases, CV with 36 cases, and AS with 10 cases. Conversely, AC, AmC and ILP exhibit the lowest levels of utilisation, each with only one documented use case, as shown in Table 4.

Table 5 presents the local government with three or more documented use-cases of AI technologies. Notably, the top seven positions are occupied by local governments in China, with Changsha leading with 11 cases, followed by Hangzhou with nine cases, and Shenzhen with seven cases. Additionally, 22 local governments are recorded with two cases, while 128 local governments have one documented case each.

Analysing the distribution of AI utilisation cases in local governments by country, it is observed that more than half of the cases originate from China, the US, and the UK, totalling 160 cases. Specifically, China leads with 62 cases, followed by the US with 53 cases, and the UK is in third place with 45 cases. Australia, Sweden, and Canada also exhibit a notable presence with 22, 13, and 12 cases, respectively (Figure 3). Furthermore, the utilisation of AI technologies varies across countries. In the US, NLP emerges as the predominant AI technology with 33 cases recorded in local governments. Conversely, the UK exhibits a balanced adoption of RPA and NLP, totalling 17 cases. Notably, China predominantly utilises NN with 25 cases and CV with 16 cases as the primary AI technologies employed within its local governments (Figure 4).

The earliest use-case was from the year 2004. However, not many use-cases were documented between 2004-2010 and 2010-2014. But a steady increase in recorded cases is observed from 2014 onwards, with a doubling of documented cases between 2017 and 2018, signifying the onset of an exponential trend of growth. Despite a slight decline in 2019 with 22 cases, there was a significant surge in 2020, with 58 cases recorded. The peak in recorded cases occurred in 2021, totalling 68 cases (Figure 5).

In terms of technology adoption trends, NLP has demonstrated consistent operational usage in local government settings from 2004 onwards, maintaining its presence until 2024. Conversely, the adoption of other technologies exhibits fluctuating growth patterns. According to recorded cases, RPA was introduced in 2015, CV and AS in 2016, and NN in 2017. Additionally, AC was introduced in 2016 and ILP and AmC in 2019 (Figure 6). It is interesting that the number of cases is dropping off across the board from 2021-2024. One possible reason behind this could be a lag behind the publication of such AI initiatives from local government.

As previously mentioned, a total of 28 sub-service categories have been documented under five main services within the dataset. Notably, Information Management emerges as the most prevalent area with 49 recorded cases, followed by Back-office Work with 33 cases, Transportation and Traffic Management with 27 cases, and Public Health with 25 cases. It is noteworthy that local governments are actively adopting AI technology across a wide array of domains, reflecting a concerted effort towards enhancing service delivery effectiveness and efficiency (Table 6).

Figure 7 illustrates the distribution of years and services within the dataset. It is evident that certain years are associated with specific services. For instance, in 2018, four main branches are connected, namely back-office work, information management, permits granting and licensing, and transportation and traffic management. Similarly, in 2020, connections are observed primarily towards three main services, namely back-office work, information management, and public health. Furthermore, in 2021, links are established towards back-office work, information management, public health, and waste collection and management. Lastly, in 2022, connections are noted towards back-office work, information management, transportation and traffic management, and waste collection and management.

4.2. AI Technology and Service Distribution

4.2.1. Natural Language Processing in Local Governments

The results of the analysis indicate that NLP has been connected towards nearly 18 services in local government. Among them, information management, back-office work, posting complaints, interpretation, and public health are the highly provided services. NLP is the technology that has been used in chatbots to understand natural human language communication (Figure 8).

These capabilities of NLP provided local government with a more effective and efficient way of:

▪

Removing language barriers—Phoenix Council, US utilises Amazon Web Services (AWS) Lex chatbot to create a conversation interface in both English and Spanish (AZ Business Magazine, 2021);

▪

Freeing up human time from performing repetitive boring tasks—Lewes and Eastbourne Council in the UK employ ELLIS, covering over 1,000 council topics and which was trained on 12,000 resident questions. It has already enabled the relocation of 5 full-time contact agents away from live chat to focus on the more complicated tasks (Govlaunch, 2022);

▪

Connecting residents to city council services 24 hours a day—The Public Relations Office within Municipality of Grosseto in Italy implemented digital functions to enhance communication between residents and the administration. A virtual assistant is available 24/7 to guide residents through online procedures and assist with problem-solving (Municipality of Grosseto, 2021);

▪

Enhancing wide-scale customer experience - The municipalities of Kortrijk, Tournai, and Roubaix collaborated to create the free Tripster chat tourism service, an overarching approach to promoting cross-border tourism and making it more accessible to everyone (Tripster, 2021).

Additionally, NLP has played a key role during the COVID-19 pandemic. The Kolkata Municipality in India used an innovative chatbot tool to streamline the vaccination process as it fights COVID-19. Within ten weeks of the chatbot-embed platform launch in mid-May 2021, it attracted more than 250,000 unique users and booked more than 75,000 vaccination appointments directly through the platform. Initially, the platform was only connected to three vaccination centres. However, it was rapidly expanded to include approximately 100 vaccination sites across the city, which significantly improved the accessibility and efficiency of the vaccination campaign (Bhatia, 2021). Furthermore, some municipalities have leveraged NLP for short-term purposes, such as during elections. For instance, the Hamilton Council in Canada introduced two innovative online tools—a voice query directory and a virtual assistant—aimed at enabling residents to effortlessly locate and access information about the municipal election held on October 24, 2022 (Hamilton City Council, 2022).

4.2.2. Neural Networks in Local Governments

NNs are integrated with 19 services and predominantly linked to transportation and traffic management, public safety and security, information management and public health. However, NNs are recognised as a computationally expensive technology due to their demand for significant processing power and time. Consequently, they have been utilised for complex service provision such as transportation and traffic management and public safety and security.

Several municipalities utilise NNs for different types of transportation and traffic management, including:

▪

Mapping ideal locations for electric vehicle charging points—Implemented by Irving Municipality in the US (VOLTA NEWS, 2023).

▪

Junction improvements—Undertaken by Lancashire County Council in the UK (Say, 2022).

▪

Determining the safest route—Implemented by Los Angeles City Council in the US (Fast Company, 2017).

▪

Assisting citizens in emergency situations such as bridge collapse—Utilised by Atlanta City Council, US (Statescoop, 2017).

▪

Analysing traffic patterns of different mode of transportation—Implemented by Kansas City Council in the US (Route-Fifty, 2017).

▪

Navigating parking system—Utilised by Hangzhou Municipality (Ascend Editorial Team, 2022), among others.

Moreover, NNs have also been utilised for safety and security, including:

▪

Predicting crime locations—Chicago Municipality adopted a model to predict when and where violent crimes are likely to occur. The former mayor Rahm Emanuel announced in early 2018 that gun violence was down 25% compared to the previous year (Emanuel, 2018).

▪

Predicting child abuse—Implemented by Hackney Council in the UK (Marsh, 2019).

▪

Safeguarding against cybersecurity issues—Utilised by Gilbert Town Council in the US (Diaz, 2022).

▪

Identifying and addressing anti-social behaviours—Undertaken by Sunderland City Council in the UK (Wray, 2022).

4.2.3. Robotic Process Automation in Local Governments

RPA is connected with 13 services, with 51 recorded cases of its usage. Among these cases, the majority were recorded for back-office work. The back office is often referred to as the engine room of the organisation (Anagnoste, 2017), where much of the work performed determines the overall success of operations (Vignesh et al., 2016; Makridakis, 2017). Its tasks encompass procurement, finance and accounting, human resource management, payroll, work reporting, and more (Bekkers, 2007; Paagman et al., 2015; Anagnoste, 2017).

Local government back-office employees often find themselves engaged in repetitive tasks for many hours each day, which can lead to an acceleration of errors and a slowdown in progress (David et al., 2023; Senadheera et al., 2024). In addition, inefficient legacy systems contribute to the accumulation of numerous pre-approval documents on desks for extended periods, resulting in suboptimal service delivery. The pre sence of outdated and inefficient administrative processes not only hampers the speed at which local governments can respond to the needs of residents but also affects the overall quality of services provided (Olowu, 2003; Ehsan, 2020).

RPA is recommended technology for back-office work, and this finding is further justified in the local government sector as well (El-Gharib & Amyot, 2023). It offers a non-invasive and cost-effective solution (Ansari et al., 2019; Plattfaut & Borghoff, 2022), which is particularly important for local councils operating in high paced work environments with limited budgets (Wewerka & Reichert, 2021; Ray et al., 2023).

In local government back-office work, RPA is utilised for various tasks including:

▪

Payslip account management, including Council’s payslip archiving system—Implemented by Surrey County Council in the UK (Surrey County Council, 2018);

▪

Management of financial assistance processing—Undertaken by Strängnäs Municipality in Sweden (Strängnäs Municipality, 2019);

▪

Validation of Blue Badge applications and invoice processing—Managed by Cumbria County Council, UK (UK Authority, 2022);

▪

Financial auditing and risk management mitigation—Liverpool City Council, Australia (Zinnov, 2023);

▪

Mileage calculations & value added tax (VAT) calculations—Handled by Gloucestershire County Council (Gloucestershire County Council, 2018).

Tax calculation (Norfolk County Council, 2021), water and sewerage service (Blacktown City Council, 2020), waste collection (Alcorcón City Council, 2022) and river management (Shanghai City Council, 2018) are among other main services provided by RPA in local government.

4.2.4. Computer Vision in Local Governments

According to the findings, CV has been employed across 13 services in local government. Among these, CV is predominantly used for transportation and traffic management as well as waste collection and management services. Some local governments have also utilised CV for transportation management, gradually extending its application to public safety. For instance, Seoul Municipality fixed CCTV on every street, which transmit data to the Transportation Operation and Information Service (TOPIS). The TOPIS website utilises this data for real-time traffic monitoring. Additionally, instances of illegal vehicle driving or parking are detected, leading to automatic fines. In cases of accidents or road constructions, detour routes are suggested, and accident notices are promptly sent to connected police and hospitals (Bandopadhyay, 2019). Moreover, there are local government AI use-cases in remote sensing to count pools, assess rooftop solar panels, electrical infrastructure asset management, leak detection in water management, and in the security/surveillance space.

Some local governments have installed fixed cameras to the municipal garbage tracks to identify:

▪

Roadside assets maintenance—Brimbank City Council, Australia (Australian Research Council, 2023);

▪

Pothole detection—Helsingborg Municipality, Sweden (Hornblad, 2022);

▪

Identification of blighted areas—Tuscaloosa Municipality, the US (Sanchez, 2021).

4.2.5. Other AI Technologies in Local Governments

Autonomous systems are used for five types of services, primarily focusing on information management in local government. For instance, Ogaki City Council in Japan use robots to guide the people to the appropriate information window or assisting them in filling out government forms (Japan Times, 2019). Additionally, affective computing, ambient computing (Bibri 2015a, b), and ILP are used for each service of local government. Affective computing is used for permit granting and licensing by London Borough Council in the UK (Davies, 2016). ILP is employed to codify building regulations in California municipalities (Krueger et al., 2019). Besides, ambient computing is utilised for public safety and security in Australia (Ku-ring-gai Council, 2019).

5.1. Why Have NLP And RPA Gained Popularity in Local Governments, and How Can these Technologies Address Specific Challenges?

NLP and RPA have gained popularity in local governments due to their profound impact on efficiency and communication. The adoption of NLP significantly enhances the quality of interaction between governments and citizens in the administrative processes/governmental affairs procedures (Androutsopoulou et al., 2019; Wang et al., 2022; Jiang et al., 2023). The common application is the NLP-driven chatbots and virtual assistants that are operational round-the-clock, efficiently addressing frequently asked questions (FAQs) and navigating users through intricate administrative procedures (Androutsopoulou et al., 2019; Cortés-Cediel et al., 2023). Chowdhary (2020) identified nine applications of NLP: indexing and searching large texts, information retrieval, classification of text into categories, information extraction, automatic language translation, automatic summarisation of texts, question-answering, knowledge acquisition, and text generation/dialogues.

These automations streamline information access and service requests and contribute to boosting citizen satisfaction with the government’s service experiences (Zhu et al., 2022; Ju et al., 2023; Sienkiewicz-Małyjurek, 2023). Furthermore, NLP’s capability to sift through and analyse copious amounts of textual data accrued in administrative processes is invaluable (Androutsopoulou et al., 2019; Nicolas et al., 2021). This process facilitates the extraction of pertinent insights, the discernment of trends, and informs governmental policy decisions (Chen & Wei, 2023; Jiang et al., 2023; Yigitcanlar et al., 2023a). An exemplar of its application is sentiment analysis, which can efficiently evaluate public opinion on diverse issues, thus empowering governments to tailor their responses to citizen concerns more efficiently while enhancing the government’s response level (Lu et al., 2023). This refined approach to public engagement and data analysis underscores NLP’s pivotal role in modernising and optimising government-citizen interactions (Wang et al., 2022; Cortés-Cediel et al., 2023).

RPA has become a crucial instrument in streamlining the routine tasks prevalent in local government operations, particularly in the context of implementing digital transformation initiatives within smart city strategies (Sobczak & Ziora, 2021). RPA’s ability to automate paperwork and repetitive administrative functions significantly alleviates staff workload (Hyun et al., 2021; Johansson et al., 2022). This automation translates into expedited processing for various administrative processes/governmental affairs procedures, such as license renewals and application processing, enhancing overall service delivery (Ranerup & Henriksen, 2019). A key benefit of RPA is its potential for cost reduction, a critical factor for local governments operating within stringent budgetary constraints (Adamczyk et al., 2021).

Additionally, the precision of automated processes reduces the likelihood of errors, ensuring heightened accuracy in data management and regulatory compliance—a vital consideration in government operations where inaccuracies can lead to substantial repercussions (Hyun et al., 2021; Sobczak & Ziora, 2021). Thus, the adoption of NLP and RPA technologies offers local governments a pathway to not only bolster operational efficiency and service quality but also promote government decisions to adapt more responsively to the dynamic needs and expectations of their citizenry, e.g., unbiased decisions, new forms of democratic participation, inclusion of users and improved working conditions for employees (Wirtz & Müller, 2019; Johansson et al., 2022; Hujran, et al., 2023).

5.2. Which Service Areas Are Most Affected by AI Technology in Local Governments, and How Does AI Improve the Efficiency in these Service Areas?

AI technology is revolutionising various service areas within local government, with significant impacts observed in the areas of administrative services, healthcare and wellbeing, transportation and urban planning, environmental management and public safety and law enforcement:

Firstly, AI facilitates the automation of routine administrative tasks in local government operations, such as data entry, document processing, and handling customer service inquiries (Ng et al., 2021). AI-powered optical character recognition (OCR) technology automates data entry by extracting information from documents such as forms, applications, and records (Baviskar et al., 2021; Sharma et al., 2022). This process reduces manual data entry errors and speeds up processing times for various administrative tasks (Pencheva et al., 2018). NLP algorithms enable chatbots and virtual assistants to handle citizen queries effectively, reducing response times (Engin & Treleaven, 2019; Susar & Aquaro, 2019). AI-based workflow automation platforms streamline administrative processes by routing tasks, assigning priorities, and automating notifications and approvals (Chen et al., 2023; Gill et al., 2024; Licardo et al., 2024). These systems optimise task management, reduce bottlenecks, and ensure smoother coordination among local government officials, thereby leading to heightened productivity and efficiency (Habbal et al., 2024).

Secondly, AI transforms transportation and urban planning by optimising traffic flow, enhancing public transit, implementing smart parking solutions, refining infrastructure planning, integrating micro-mobility options, and promoting transportation equity (Kamrowska-Załuska, 2021; Paiva et al., 2021; Mitieka et al., 2023). Smart traffic management systems use AI to analyse real-time traffic patterns and accordingly adjust signal timings to improve flow and reduce congestion (Araghi et al., 2015). Urban planners utilise AI to simulate scenarios for infrastructure development, optimising city layouts for improved mobility (Alahi et al., 2023; Bibri et al., 2024; Son et al., 2023). These AI-driven solutions improve efficiency, reduce congestion, minimise emissions, and foster inclusive urban development, creating more sustainable and resilient local government functions (Kamrowska-Załuska, 2021; Son et al., 2023).

Thirdly, in local government, AI revolutionises environmental management by tracking pollution sources, optimising waste management, enhancing energy efficiency, conserving natural resources, aiding climate adaptation, and supporting emergency response (Andeobu et al., 2022; Bibri et al., 2023a, b). By leveraging AI across these fronts, local governments enact proactive measures, reduce environmental impact, and promote sustainability, safeguarding communities and fostering resilience for future generations (Buyya et al., 2018).

AI-powered chatbots and virtual assistants interact with citizens, providing information on healthcare services, scheduling appointments, and offering initial triage for medical inquiries (Agarwal et al., 2022; Senadheera et al., 2024). This reduces the burden on staff, improves accessibility to services, and ensures timely assistance for citizens. The COVID-19 pandemic has prompted local government entities to significantly utilise AI technologies to bolster health and safety measures (Costa & Peixoto, 2020). AI algorithms analyse health data and social media to detect outbreaks, track trends, and predict risks, enabling targeted interventions and resource allocation to mitigate disease spread (Elavarasan & Pugazhendhi, 2020; Agarwal et al., 2022).

Finally, AI-powered chatbots and virtual assistants engage with communities, offering crime prevention tips, safety information (Blauth et al., 2022), and aid in reporting incidents, fostering trust, transparency, and collaboration between local government and citizens for more effective emergency services (Alahi et al., 2023). AI systems prioritise emergency calls, assess incident severity, and recommend response strategies, enhancing dispatch efficiency, reducing response times, and optimising resource allocation during emergencies (Farahani et al., 2020; Costa et al., 2022).

In summary, local governments can significantly enhance operational efficiency by leveraging AI to automate tasks, analyse data, predict trends, enhance citizen engagement, and optimise resource allocation. This enhanced efficiency facilitates more effective public service delivery, enabling the tackling of complex challenges with greater precision and agility. The integration of AI technologies into the operational frameworks of local governments not only streamlines administrative processes but also enriches the decision-making ecosystem with data-driven insights. This technological empowerment facilitates a more responsive, transparent, and inclusive governance model and contributes to fostering a more connected and satisfied community.

5.3. Why Do Public Safety and Law Enforcement Get Less Attention on Local Government AI Applications?

Local governments often prioritise AI applications that directly address immediate challenges or offer clear efficiency gains in service delivery (Mikhaylov et al., 2018; Dwivedi et al., 2021). The results indicated that AI initiatives in areas such as administrative service, transportation, and urban planning may receive more attention than public safety. Meanwhile, public safety issues are often multifaceted and sensitive, involving considerations of law enforcement, emergency response, community relations, and privacy rights (van Dijk et al., 2019; Nassar & Kamal, 2021).

Implementing AI solutions in this domain requires careful planning, stakeholder engagement, and consideration of ethical and legal implications, which can pose challenges for local government AI projects (Gentzel, 2021). Local governments often face resources in terms of both funding and expertise when it comes to developing and implementing AI solutions for enhancing public safety (Mikhaylov et al., 2018; Gracias et al., 2023). Consequently, they prioritise their investments based on immediate needs and may allocate resources to other services. For instance, Seoul Municipality initially installed CCTV cameras for real-time traffic monitoring. However, during emergencies, they repurpose these cameras to enhance city safety, demonstrating a flexible and adaptive approach to resource utilisation.

In the newsletter of Forbes, Dan Hoffman, the city manager of City of Winchester, stated that “…in time though, I see AI having an impact as big, if not bigger, in public safety. However, those will take longer to develop and trust for obvious reasons. Adoption in the first responder community will also take time as those systems are traditionally more expensive and require more standards and training. However, when the time comes, there will be huge leaps in AI tools that help prevent fires and medical emergencies. We’re already seeing AI tools employed by groups like the National Centre for Missing and Exploited Children making great strides in fighting child sex abuse. So, we’re seeing larger national organisations use AI for public safety; it’s just a matter of time before it is more common at the local level” (Schmelzer, 2020).

Furthermore, addressing the concerns and challenges associated with implementing AI in public safety necessitates a long-term strategy incorporating input from a broad array of stakeholders and establishing ethical guidelines. This strategy is required to guarantee that AI technologies are deployed in a manner that is transparent, responsible, and effective, especially within community settings. As Fukuda-Parr & Gibbons (2021) suggested, AI-driven application in public safety requires a deliberate and considered framework that balances technological innovation with ethical considerations, safeguarding individuals’ and communities’ rights and well-being while enhancing public safety measures.

5.4. Why Do Different Local Governments Use Different AI Systems?

The findings reveal a diverse landscape where different local governments employ various AI systems for service provision. This diversity stems from the unique needs and priorities of each locality, shaped by factors like population size, geographic location, economic activity, and social demographics (Brownson et al., 2009; Tan & Taeihagh, 2020). For instance, a city prone to natural disasters may prioritise AI systems for disaster prediction and prevention, while an urban area may focus on urban issues such as traffic optimisation. The selection of specific AI systems is influenced by a myriad of factors (Duan et al., 2019), including budgetary constraints, availability of technological infrastructure, and human capital (Thuan et al., 2015; Votto et al., 2021; Kelly et al., 2022). Local governments with robust resource allocation mechanisms tend to leverage AI systems more effectively and efficiently (Păvăloaia & Necula, 2023).

Furthermore, each local government grapples with its unique set of challenges, spanning from crime rates and environmental concerns to transportation congestion (Madumo, 2015; Mikalef et al., 2019). In response, AI systems are tailored to address these specific challenges, resulting in varied deployment strategies based on local needs (Peres et al., 2020). As primary service providers to the community, local governments must accommodate the preferences of a diverse range of stakeholders, including community members themselves (Beynon-Davies & Martin, 2004). However, aligning these preferences can be complex, as public officials, community leaders, and citizen stakeholders may harbor differing priorities and opinions regarding AI adoption (Miller, 2022). Moreover, local governance structures and political dynamics significantly influence the selection and implementation of AI systems (Dafoe, 2017).

Overall, the adoption of various AI systems across different local governments is driven by a multitude of factors, including distinct needs, resources, regulations, and stakeholder preferences. This diversity necessitates thorough evaluation by each local government to select AI solutions that best align with their specific challenges and objectives. It is imperative for AI deployment in local governments to not only address the unique needs of communities but also to comply with local regulatory frameworks and to accommodate the expectations of diverse stakeholders. Tailoring AI deployment in this matter can enhance the capacity of local governments for service delivery, informed decision-making, and overall operational efficiency, particularly within specific community contexts.

5.5. What Are the Key Challenges, Future Impacts, and Trends?

The adoption of AI technology in local government holds promise for enhanced efficiency and service delivery, but also poses significant challenges that demand deliberate, careful consideration and strategic planning. One major hurdle lies in integrating AI technologies with existing legacy systems (Sun & Medaglia, 2019; Pencheva et al., 2020; Dwivedi et al., 2021), as these systems may not be designed for compatibility with AI. This integration process often requires extensive overhauls or replacements and entails seamless functionality and communication between new and established systems (Wang et al., 2021; Dwivedi et al., 2021). Additionally, since AI-driven applications rely heavily on data, safeguarding the privacy and security of sensitive information managed by governments becomes a critical concern (Sun & Medaglia, 2019; Alshahrani et al., 2022; Sienkiewicz-Małyjurek, 2023).

This challenge includes considerations not only of the technical aspects of data security but also of the ethical dimensions and the preservation of public trust (Aoki, 2020; Chen et al., 2021; Ingrams et al., 2022). Moreover, the potential of AI algorithms to unintentionally reflect and amplify biases present in their training data necessitates vigilant oversight of the ethical repercussions of AI-driven decision-making (Sunarti et al., 2021; Shaamala et al., 2024). This oversight is imperative to ensure equitable and non-discriminatory application and practice of these technologies (Li et al., 2023a; Saeed & Omlin, 2023). Another pivotal challenge is navigating the evolving regulatory landscape, particularly concerning data usage and privacy. Governmental entities are compelled to ensure adherence to extant laws while remaining adaptable to the dynamic regulatory context (Sun & Medaglia, 2019; Campion et al., 2022; Li et al., 2023b). As regulations governing data privacy and use continue to evolve, local governments must stay informed of these changes and adapt their practices accordingly. This adaptation is critical to ensure compliance with legal standards and uphold ethical principles in the implementation of AI technologies.

Furthermore, the financial consideration of implementing AI solutions may pose a significant barrier, especially given the budgetary constraints typical of local governments (Wang et al., 2021; Yigitcanlar et al., 2023b). In this context, it is imperative for local governments to meticulously evaluate the balance between financial outlays and the potential long-term advantages of AI implementation. This evaluation must prioritise ensuring that the realised benefits align not only with the initial promises of AI adoption and utilisation but also substantively contribute to enhancing the well-being of citizens (Nzobonimpa & Savard, 2023). It is critical that these advancements are achieved without inadvertently compromising the welfare of any community segments, thereby maintaining a holistic approach to public service enhancement through AI integration (Williams et al., 2016; Jang, 2023; Mesa, 2023).

Besides, the deployment of AI technologies requires a workforce with specialised skills, not only in terms of recruiting new talent but also in training existing employees to operate and manage these advanced systems competently (Wang et al., 2021; Mikalef et al., 2022). This need for skill development may represent a time-intensive and financially demanding undertaking, adding to the complexity of adopting AI in local government. It underscores the importance of investing in human capital development alongside technological infrastructure to ensure the successful deployment and sustainable functioning of AI solutions within local government contexts.

Lastly, it is important to note local government cybersecurity (Hossain et al., 2024), and the role of fake/synthetic data/poisoned data and adversarial attacks, and how these may affect services provided through AI enabled local government systems are among the factors to be carefully considered.

5.6. Limitations of the Study

It is imperative to recognise the constraints that could affect how the results are interpreted: (a) This study is purely based on a grey literature review, which often lacks the rigorous peer-review process of academic literature, leading to variations in quality and reliability; (b) The analysis results presented here could involve unintended bias of authors; (c) There is a possibility that there exist more than 262 real-world use-cases; however, our research methodology may not have been able to locate them, potentially resulting in gaps or the introduction of biases into the data synthesis process; (d) The study encountered challenges in accessing Chinese cases, which required additional efforts for data collection. This entailed conducting a secondary search using Baidu for rectification and translation purposes, a process which was labour-intensive; (e) It is important to acknowledge the shortcomings in the available case information—Each piece of literature reviewed presented large differences across their format and the level of detail presented, ranging from comprehensive descriptions to brief overviews. As a result, the key characteristics of the cases were restricted to AI technology, local government name, country name, comprehensive use description, and year of introduction; (f) Today in many organisations, people are using Generative AI—such as ChatGPT—to perform routine office tasks without policy guidance or oversight or ‘under the radar’. This is almost certainly true in local governments too, especially in citizen facing roles, but will not show up in an online trawl, but certainly represents an important local government use case. Nonetheless, despite these limitations, the study was carried out in a measured and transparent manner, aiming to provide valuable insights into AI adoption in local governance while considering these gaps in the methodology and data sources.