Using AI in Performance Management: A Global Analysis of Local Government Practices

Abstract

The integration of artificial intelligence plays a critical role in human resource management in local governments by ensuring smooth, essential HR operations, including recruitment, performance management, and workforce planning. The current study is a systematic review focused on determining the performance management factors that should be considered when using artificial intelligence in the local government sector. Although artificial intelligence (AI) is becoming increasingly integrated into the governance and administrative systems of local governments around the world, this study raises critical questions about how performance should be managed, measured, and improved. Articles were screened based on their title, abstract, and keywords, following which the inclusion and exclusion criteria were applied. A comprehensive search was conducted in the EBSCOhost, Emerald Insight, Taylor & Francis, Scopus, and SpringerLink databases. These databases were chosen because they are prominent sources that publish various materials related to the social sciences. This scoping review followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines and included 22 peer-reviewed empirical studies published between 2015 and 2025. Analysis of the identified 22 peer-reviewed articles revealed that the successful application of AI in local government performance management depends on six critical performance management factors: data quality and accessibility; strategic alignment with performance goals; evaluation criteria and metrics; ethical and legal oversight; institutional capacity and leadership; and change management and stakeholder engagement. These factors are interdependent and represent both technical and organisational dimensions of public administration. This study highlights that AI entails more than innovation; it reshapes the foundations of performance governance, requiring new capabilities, values, and institutional practices.

1. Introduction

As organisations are preparing to remain competitive in the digital age, it is essential to acknowledge that the traditional framework of performance management—particularly employee evaluations—is seeing a significant transformation, propelled by the capabilities and insights provided by artificial intelligence (AI) technologies (Lee et al., 2025). AI is a cutting-edge technology that is rapidly evolving and can be employed to enhance human resource capabilities in the modern workplace in the era of swift digital change (Bankar & Shukla, 2023). The implementation of artificial intelligence in human resource management has the capacity to transform essential HR processes, including recruitment, performance management, and workforce planning, by employing sophisticated algorithms to improve accuracy and efficiency (Taslim et al., 2025). Artificial intelligence technologies are revolutionising traditional performance management approaches, facilitating more efficient and data-informed decision-making (Nath et al., 2025).

The process of organising, monitoring, and enhancing the performance of people, groups, and the company is known as PM. Increasing productivity, setting performance goals, evaluating performance, providing feedback, performance development, and the facilitation of sound decision-making are the objectives of PM (Faozen & Sandy, 2024). The process of PM, which involves managers and staff working together to create goals, evaluate outcomes, and eventually recognise exceptional performance, has a large impact on organisational success (Modika et al., 2023). By its very nature, PM should increase organisational effectiveness and assist the organisation in accomplishing both its strategic and operational goals if it is appropriately implemented, analysed, and evaluated (Ngubane, 2022). PM is crucial in local governments not only as a management and communication tool, but also as a source of budgetary information (Modika et al., 2023). It is important to note that performance management in this study is interpreted broadly, encompassing both the evaluation and enhancement of individual employee performance and the broader strategic and operational performance of local government organisations and their policies.

However, despite the existence of PM, according to a study conducted by Sorano et al. (2023), local government is seen as complex from the organisational perspective, with several aspects influencing PMS design. Local governments around the world face significant obstacles in successfully implementing a PMS due to a range of variables, such as regulatory changes, a lack of experience, and a lack of integration between the planning and budgeting processes (Pudjono et al., 2025).

Employers are increasingly focusing on employee performance management as a technique for enhancing resource allocation and utilisation efficiency in local governments around the world (Mohangi & Nyika, 2023). To remain competitive, local governments are required to enhance their operational efficiency and performance by providing better services to citizens. Globally, local governments are legally required to support seamless service delivery, demonstrating organisational performance that satisfies residents’ expectations and goals for quality of life (Draai & Zazi, 2021). Without a doubt, local government is the most significant domain of government, as it operates at the grassroots level and is responsible for providing crucial services to citizens at the local level (Nzama et al., 2023; Modika et al., 2023). The role of the local government is significant because it represents the most accessible form of governance to the people (Schoeman & Chakwizira, 2023).

Local government, as the governmental tier closest to citizens, has long been regarded as a failed sphere of government as it has failed to provide enough services to its inhabitants. Modika et al. (2023) echoed this sentiment, pointing out that certain municipal governments continue to operate poorly. According to recent research, inadequate local government service delivery results from local government employees’ subpar performance (Masiya et al., 2021). The research objective for this study was to determine the performance management factors that should be considered when using artificial intelligence in the local government sector. Therefore, local governments should be committed to collaborating with individuals and community groups to identify long-term solutions to address their social, economic, and material requirements while also improving their quality of life (Masiya et al., 2021). Citizens also want their government to be efficient and accountable for everything they do and ensure that their needs are addressed (Beshi & Kaur, 2020; Schoeman & Chakwizira, 2023). The implementation of PM could assist local governments in holding their employees accountable for poor performance, which results in the delivery of poor services to the citizens. Employers and employees can explicitly state an organisation’s goals and how they must be achieved using a PMS (Ndasana & Umejesi, 2022). This study followed a systematic review approach, and a comprehensive search was conducted in the EBSCOhost, Emerald Insight, Taylor & Francis, Scopus, and SpringerLink databases. The article is structured into four sections. Section 1 of the article provided an overview of the existing knowledge in relation to the research topic. Section 2 then describes the materials and methods used to collect data. Section 3 contains the study results, followed by a discussion of the findings. The final portion gives the study’s findings, limitations, and future directions.

1.1. Theoretical Framework

This study is grounded on the New Public Management (NPM) Theory and the Technology Acceptance Model (TAM).

New Public Management (NPM) Theory: The previous PM model has faced significant criticism for failing to provide the populace with commodities and services (Islam, 2015). New Public Management (NPM) was launched in the Western world, which was then adopted in the southern African continent and Sub-Saharan Africa to reform the public sector and its approaches to managing state affairs (Munzhedzi, 2021). Indahsari and Raharja (2020) defined New Public Management as performance-oriented rather than policy-oriented. It seeks to increase public sector organisations’ efficiency, effectiveness, and service quality by implementing market-based methodologies and performance-oriented management strategies (Abdullahi, 2024). NPM aims to improve the public sector’s efficiency within the parameters of private sector logic, including an understanding of the market and competition, a customer focus, performance standards, and decentralisation (Terrance & Uwizeyimana, 2023). From this perspective, NPM promotes efficiency, accountability, and performance, as well as approaches commonly used in the commercial sector, such as outcomes-based performance evaluations. The NPM theory will play a fundamental role and is more applicable to our study since it helps explain why and how local governments around the world use AI to improve service delivery and performance results. NPM’s emphasis on “efficiency, effectiveness, and service quality by implementing market-based methodologies and performance-oriented management strategies” directly aligns with the drivers for AI adoption in local government to enhance performance. It provides a historical and theoretical context for the shift towards performance-oriented governance in the public sector, which sets the stage for understanding the importance of the identified performance management factors for AI’s role. Building upon the principles of New Public Management (NPM), which emphasise efficiency, accountability, and performance-oriented strategies in public administration, understanding performance management in local governments is critical, especially given the transformative potential of artificial intelligence.

Technology Acceptance Model (TAM): This study is also based on the TAM theory, which explains how people adopt and use technology (Musa et al., 2024). According to the TAM, users’ attitudes and intentions toward technology are influenced by their assessment of its usefulness and simplicity of use (Aljarrah et al., 2016). According to Tambun et al. (2020), the TAM theory consists of four aspects:

• User behaviour, which is the actual operational behaviour of users of the new technology;
• Behavioural intention, which is the willingness of users to try new technologies;
• Perceived usefulness, which is the users’ subjective understanding of the utility of the newly adopted technology;
• Perceived ease of use, which is the degree to which technology users make use of new technologies.

The widespread adoption of AI in local government necessitated a rethinking of traditional PM approaches. The rise of AI is critical for local government progress and long-term service delivery. Therefore, researching the potential of AI in PM is critical as its successful implementation demands a high level of acceptance of digital technology (Schorr & Gorovoj, 2023). This theory is particularly applicable to this study since it examines how employees accept and embrace new technology in the local government. At the same time, it makes a useful contribution by detailing how employees in the organisation will be confident and willing to include AI in PM.

1.2. Theoretical Perspectives on Performance Management

Employees are critical to the success of any organisation since they are the primary resource in ensuring that strategic actions are carried out with due diligence and the appropriate timing (Mphahlele & Dachapalli, 2022). Their employment will benefit the organisation only if their performance is monitored and evaluated in order to achieve the organisation’s aims. The term performance management is defined differently by several researchers based on different situations (Modika et al., 2023). In this study, performance management in local governments will be discussed. PM primarily involves overseeing employees and organisational operations; if inadequately managed, it can adversely affect employee performance by diminishing job satisfaction, lowering morale, and consequently hindering productivity (Simpson & Simpson, 2022). PM can be viewed as providing employees with performance feedback (Marie & Khumalo, 2024).

PM is not an instrument but an ongoing process, which focuses on planning, assessing, and offering support for improved performance, rather than penalising underperformance, while also rewarding good performance (Van Waeyenberg et al., 2022). For this reason, it is imperative that organisations continuously assess their employees’ performance in order to understand their existing and future capabilities (Samwel, 2018). PM tries to identify the organisation’s goals, the tactics required to accomplish them, how to implement these strategies, and who will be the driving forces behind these changes. PM focuses on how an organisation can achieve its goals in an effective manner (Simpson & Simpson, 2022). It is a systematic procedure that improves organisational performance by developing the performance of individuals and teams (Rameshbabu, 2017).

1.3. AI in Performance Management Factors

The organisation’s success and growth depend on efficient employee performance management (PM) processes. In recent years, artificial intelligence (AI) has emerged as a transformative technology in the human resources sector, offering substantial advantages. AI is swiftly revolutionising human resource management (HRM) by improving the efficiency and efficacy of essential operations, including recruiting and PMS operations (El-Ghoul et al., 2024). AI can facilitate the adoption of innovative methodologies and tools that provide more dynamic, objective, and data-driven strategies for managing and improving employee performance (Thirunagalingam et al., 2025). Its integration can transform the landscape of human resource operations, particularly in PM, to improve how organisations evaluate their employees and track their performance (Khanna, 2025). With the digital age’s rapid and unpredictable evolution, integrating AI into PM will provide the organisation with useful insights for critical decision-making processes such as promotions, merit raises, transfers, and training and development (Basnet, 2024).

AI in the workplace is transforming several aspects of modern business, including PM (Thirunagalingam et al., 2025; Khanna, 2025), providing firms with strong capabilities for tracking, analysing, and improving employee performance in real-time (Alrakhawi et al., 2024). Traditionally, human resource processes such as personnel selection, appraisal, and recruiting were carried out manually, which could be biased in some situations. AI solutions are required for PM analysis of development and HRM competencies (Gaol, 2021). Through the use of AI, in the future, PM could be inextricably tied to real-time feedback, predictive analytics, and tailored development paths (Nyathani, 2023).

The widespread adoption of AI in local government necessitated a rethinking of traditional PM approaches. The rise of AI is critical for local government progress, offering both short-term and long-term benefits. In the short term, AI enables task automation, faster output, and reduced workload. In the long term, it supports organisational redesign, strategic value creation, and sustainable innovation (Mortensen, 2025). Therefore, researching the potential of AI in PM is critical as its successful implementation demands a high level of acceptance of digital technology (Schorr & Gorovoj, 2023).

2. Materials and Methods

This scoping review examines the application of artificial intelligence (AI) in performance management across local government sectors around the world. The main research question addressed in this review is “Which performance management factors must be considered when using artificial intelligence in the local government sector? This was a qualitative study; qualitative research addresses how and why issues rather than how many or how much (Tenny et al., 2017). Only secondary data was collected and analysed, which facilitated an in-depth understanding of this complex issue and offers insights into the performance management factors when using AI in local governments. This study is grounded in a constructivist epistemology, which aligns with the interpretative nature of qualitative research on local governments. Performance management within human resource management reforms in post-socialist local governments is a socially constructed process. Consequently, this research followed a constructivist ontology, assuming that reality is not universally fixed but contingent upon governance structures and policy frameworks. Based on this perspective, the study aimed to explore which performance management factors must be considered when using AI. Therefore, a qualitative approach is the most appropriate for a contextually rich and theoretically informed analysis of these factors.

2.1. Protocol and Reporting Standards

This scoping review followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines. The review protocol was developed before the data collection and adhered to a transparent and reproducible methodology throughout the search, screening, and synthesis processes. The protocol was divided into four main stages: (1) Identification, where potentially relevant records are retrieved from databases using search strings; (2) Screening, in which duplicates are removed and records are assessed against preliminary inclusion criteria based on titles and abstracts; (3) Eligibility, where full-text articles are reviewed against detailed inclusion and exclusion criteria; and (4) Inclusion, which results in the final sample of articles used in the synthesis. This staged approach ensured systematic, transparent, and replicable filtering of literature for the current study.

In this study, two reviewers independently screened all retrieved records at the title/abstract and full-text levels. Conflicts were resolved through consensus discussions to minimise reviewer bias. No automated bias detection tools were used, but peer-reviewed status and adherence to empirical design acted as quality filters. As this review was designed as a scoping review and not a meta-analysis, no formal effect measures (e.g., risk ratios, mean differences) were calculated. Instead, emphasis was placed on mapping study contributions thematically and narratively. While no formal statistical assessment of reporting bias was conducted, the broad database search (five academic databases) and transparent application of inclusion/exclusion criteria minimised the risk of missing eligible studies. Grey literature and non-English works were excluded by design. Confidence in evidence was not formally rated (e.g., GRADE not applied). Instead, certainty was inferred narratively by highlighting convergence across studies, consistency of themes, and repetition of key performance factors across multiple geographic and institutional contexts.

2.2. Information Sources and Search Strategy

A comprehensive search was conducted across the following academic databases:

• EBSCOhost;
• Emerald Insight;
• Taylor & Francis;
• Scopus;
• SpringerLink.

These databases were chosen because they are prominent sources of various materials related to the social sciences and to minimise publication bias. Heck et al. (2024) discovered that incorporating many databases in a systematic review yields the best proportion of relevant material. The inclusion of these databases ensured a robust representation of the sample. The search strategy combined Boolean operators and controlled vocabulary. The Boolean operators were used to retrieve the relevant results within the scope of the study by using keywords. The final search string was as follows: (“artificial intelligence” OR ai OR “A.I.” OR “machine learning” OR “deep learning” OR “algorithmic decision-making” OR “automation” OR “predictive analytics”) AND (“local government” OR “local authority” OR “municipality” OR “town council” OR “regional government”) AND (“performance management” OR “performance evaluation” OR “performance measurement” OR “local government performance”). These keywords were core requirements for this study to capture relevant information from existing literature. Searches were conducted in February and March 2025, and all retrieved citations were exported for screening and analysis. The articles were screened based on their title, abstract, and keywords, and subsequently, the inclusion and exclusion criteria were applied.

2.3. Inclusion and Exclusion Criteria

To ensure a methodologically rigorous selection of studies, this research employed a combination of purposive and criterion sampling, two widely recognised qualitative sampling strategies (Palinkas et al., 2015; Campbell et al., 2020). Purposive sampling is extensively used in qualitative research because it allows for the selection of information-rich cases that provide the most relevant insights into the research topic while ensuring methodological rigour (Patton, 2002). The following inclusion and exclusion criteria were used to obtain the most relevant published articles for the analysis.

The inclusion criteria were as follows:

• Empirical studies (qualitative, quantitative, or mixed methods).
• Focused on local or regional government contexts.
• Examined the use, role, or impact of artificial intelligence (AI) or related technologies (e.g., machine learning or algorithmic systems).
• Addressed performance management, performance evaluation, or decision-making processes.
• Published in a peer-reviewed journal between 2015 and 2025.
• English language.

The exclusion criteria were as follows:

• Review articles, theoretical commentaries, or editorials;
• Articles not focusing on local governments;
• Articles not addressing AI or performance-related topics;
• Non-English or grey literature.

(i) Only empirical studies were considered because they strengthened our results through their reliance on objective, observable evidence, which minimised personal bias and increased the reliability and generalisability of the findings. Unlike approaches based on anecdotes or authority, the empirical research used rigorous methods to lead to evidence-based knowledge that can be confirmed or disproved by others, ultimately advancing understanding across various fields. (ii) The rationale for restricting our systematic review to English-language studies often stemmed from practical constraints, such as time, cost, and resource limitations associated with translation and interpretation. However, this practice introduces a language bias, potentially excluding relevant research, skewing conclusions, and limiting the diversity of perspectives in the review. This exclusion can lead to an incomplete understanding of a phenomenon and may significantly impact the generalisability and reliability of the review’s findings. We will therefore add this as a limitation to the study. (iii) The rationale for a 2015–2025 time frame includes aligning with Sustainable Development Goals (which began in 2015), significant global events like the COVID-19 pandemic, and shifts in technological advancements.

2.4. Selection of Sources Process

The selection of sources followed a systematic screening process to reduce the possibility of bias and to ensure transparency. Additionally, six databases were consulted to minimise bias. All the identified records were imported into a spreadsheet for screening. The process involved the following steps:

• Title, abstract, and keyword screening to remove irrelevant studies;
• Full-text review of potentially eligible studies;
• A total of 22 articles were selected for the final analysis based on the inclusion criteria.

A PRISMA flow diagram was used to document the selection process, which included the following information:

• Number of articles retrieved per database;
• Records excluded during title/abstract screening;
• Full texts excluded with reasons;
• The final number of articles included in the synthesis.

The detailed PRISMA flow diagram is presented in Figure 1 below.

The results from each searched database were first exported. These individual files were then manually consolidated into a single master reference list. Following this merger, a rigorous manual deduplication process was conducted. This involved a careful screening of the entire list to identify and remove duplicate entries by comparing titles, authors, and publication details. To ensure accuracy, this manual screening was performed twice. No specialised citation management software was used for automated deduplication. This systematic, manual procedure was a planned part of our research protocol designed to ensure the integrity of the final dataset used for screening. To ensure the integrity and reproducibility of our dataset, the following crucial steps were taken:

(i)

Database Merging: The search results from each of the selected databases were exported individually. To create a single, unified dataset for screening, these separate reference lists were manually consolidated into one master list.

(ii)

Duplicate Removal: Once the master list was created, a manual screening process was undertaken to identify and remove duplicate entries. This involved a meticulous review of the entire list, comparing citations based on titles, authors, and publication years to ensure each unique study was represented only once. This process was conducted again to verify its accuracy. In line with the reviewer’s query, we can confirm that no specialised citation management software was used for automated deduplication; the entire process was conducted manually to ensure careful oversight.

(iii)

Planning and Reproducibility: This two-step manual procedure (consolidation followed by a rigorous two-pass screening) was an integral part of our research protocol.

This study discovered 2622 publications through the database searches, and 200 papers were vetted based on their abstracts and titles. The inclusion and exclusion criteria were then applied. A total of 123 publications were removed because they were not empirical, did not use AI, or were in an unrelated field. The authors evaluated 77 full-text articles for eligibility, and 55 were excluded because they were not empirical, did not involve AI, or were in another field. Only 22 articles were included in the qualitative synthesis.

Drawing on 22 peer-reviewed empirical studies published between 2015 and 2025, the scoping review identified key themes and performance factors that shape the integration of AI into public administration. Articles from 2015 to 2025 (a decade) were chosen to facilitate a thorough investigation of both short-term and long-term aspects, namely regarding local government adoption, which aligns with the study’s aim. The period 2015–2025 is suitable for a systematic review of performance management because it encompasses a significant era of transformation in the field, including a shift from traditional annual reviews to more continuous and agile systems, increased integration of data and analytics, a growing focus on employee development and well-being, and a greater emphasis on aligning individual contributions with organisational goals. Using the PRISMA-ScR framework, the review systematically presents its methodology, thematically synthesises the findings, and offers practical and theoretical insights for local government policy and practice. The PRISMA protocol is extensively utilised to guarantee that all facets of research are reported with precision and transparency (Sarkis-Onofre et al., 2021). (i) The PRISMA protocol was deemed the most appropriate approach because its framework directly aligns with the design of this systematic review. It provided a structured, evidence-based pathway for defining our search strategy, setting eligibility criteria, conducting the screening process, and synthesising the findings. It also acted as a “road map” to ensure the transparency and completeness of what we have done in the review and what was found. It guided us as researchers on what to report and at what level of detail. (ii) Adherence to PRISMA significantly contributes to the standardisation, comparability, and reproducibility of our results. By following this globally recognised standard, our methods are transparent and can be compared to other reviews in the field. More importantly, it provides a clear and detailed audit trail of the study selection process, which is critical for reproducibility. The PRISMA adoption fostered standardisation by providing a structured, evidence-based checklist and flow diagram that dictated the minimum reporting items for this systematic review. This consistency improved comparability by allowing us as researchers to more accurately compare the methods and findings across different systematic reviews on similar topics. It also enhanced reproducibility by ensuring that the methods used in this study were clearly and completely described, allowing other researchers to understand, evaluate, and potentially replicate the entire process. (iii) Finally, the protocol helps minimise bias by requiring the pre-specification of inclusion and exclusion criteria, ensuring the objective selection of studies. Adopting PRISMA also ensures our study meets the high editorial and academic requirements for systematic reviews, thereby enhancing its methodological rigour and credibility. The PRISMA guidelines include a checklist and flow diagrams to document the systematic review process, from study identification and selection to the final reporting. By requiring us as authors to provide clear, detailed information on our methods and decision-making, the PRISMA enabled researchers to assess any potential biases and the overall quality of the review. Furthermore, the PRISMA assisted the authors in minimising bias by promoting transparency and rigour in this systematic review through its checklist, flow diagram, and protocol. Key strategies included developing a detailed protocol with clear objectives, conducting comprehensive literature searches across various sources, employing independent reviewers for study selection and data extraction to reduce selection bias, assessing the quality of included studies to minimise bias, and transparently reporting the entire process to identify potential biases and enhance reproducibility. The review protocol was not prospectively registered, which may raise questions of transparency compared with registered reviews. However, PRISMA-ScR guidelines were followed rigorously, and inclusion/exclusion criteria were applied systematically to mitigate potential bias.

2.5. Data Extraction and Synthesis Methods

Eligibility was determined by tabulating study characteristics (authors, year, journal, context, study design, AI domain, and key findings) against inclusion/exclusion criteria. Data were then synthesised using a framework-based thematic analysis. This allowed mapping of concepts (themes and performance factors) across the included studies, which is appropriate for scoping reviews that aim to identify the breadth of evidence rather than measure effect sizes. No statistical conversions or imputations were required, as quantitative pooling was not conducted. Instead, data extraction involved structured spreadsheets where study-level details were systematically charted and coded for consistency across reviewers. Results of individual studies were presented in structured summary tables (e.g., themes for Research Question 1 and performance management factors for Research Question 2). A PRISMA flow diagram displayed the selection process, and a synthesis matrix linked codes, themes, and contributing articles. Sensitivity analyses were not conducted, as the review did not employ pooled effect sizes. Instead, robustness was reinforced through transparent reporting of inclusion/exclusion decisions and triangulation between reviewers.

The data collection and synthesis focused on collecting information regarding focused on determining the performance management factors that should be considered when using artificial intelligence in the local government sector. Codes and themes were generated through iterative full-text analysis and synthesised using a framework approach. A custom evaluation spreadsheet was developed to capture key study characteristics:

• Author(s), year, title, and journal;
• Geographic context;
• AI application domain;
• Study type (qualitative, quantitative, or mixed methods);
• Key findings and relevance to the research question.

Each article was reviewed and coded based on the emerging themes using qualitative thematic analysis. Contributions to themes were interpreted and logged in a synthesis matrix.

3. Results

No formal risk of bias tool was applied in this study, as scoping reviews do not require it. However, reliance on peer-reviewed, empirical, English-language articles between 2015–2025 acted as a safeguard. This ensured baseline quality and credibility, though geographic and contextual bias could not be fully eliminated. This is a scoping review and not a meta-analysis. Individual study contributions were mapped into broader themes and factors rather than reported as standalone effect sizes. For instance, studies were categorised under AI readiness, governance, and ethical oversight. Each thematic table illustrates how studies contributed to answering the research questions. No direct statistical test was used to detect reporting bias. However, the transparent inclusion of a PRISMA flow diagram, comprehensive search strategy, and clear exclusion reasons reduced the likelihood of selective reporting. Certainty in the evidence was expressed through patterns of convergence across diverse studies. Where multiple studies in different contexts identified similar enabling factors (e.g., data quality, leadership capacity), greater confidence was inferred. Limitations, however, were acknowledged due to uneven geographic representation and exclusion of grey literature.

This section is organised into three sections. The first section provides a Geographic distribution of the study area in articles. The second section provides the Publication Year of the reviewed articles. The third section presents the distribution by databases. The last section provides the data charting process and themes based on the number of reviewed articles.

3.1. Geographic Distribution of the Study Area in Articles

Figure 2 below presents the geographic distribution of the study area in the articles.

The geographic distribution of the study area in the articles highlights the growing global adoption of artificial intelligence (AI) in performance management (PM) within local governments. A total of 22 countries have published research in this field, reflecting the increasing international interest. As shown in Figure 2, articles contributed largely in this study are from Ghana (2), Estonia (2), Nigeria (2), Italy (2) and Bangladesh (2) as shown in Figure 2. In contrast, nine countries have produced only a small number of publications. These findings suggest that there is still significant potential for broader global engagement in this emerging area of research.

3.2. Publication Year

The comprehensive study of publishing years is presented in Figure 3.

3.3. Distribution by Databases

Figure 4 illustrates the distribution of databases used in this study.

3.4. Data Charting Process and Themes

Table 1. Comprehensive coding framework.

Code (Performance Factor)	Definition	Open Codes	Contributing Articles (Author, Year)
Data Quality and Accessibility	The extent to which local government data is accurate, timely, interoperable, and AI-ready	Data interoperability, real-time input, and accessible databases	Roztocki et al. (2023), Zhao et al. (2022), Waheduzzaman and Miah (2015), Anthony Jnr et al. (2022)
Strategic Alignment with Performance Goals	AI adoption that supports the performance vision and strategic planning of the institution	Strategic goals alignment, logic model mapping, and AI-supported strategy	Karampotsis et al. (2024), Kenk and Haldma (2019)
Evaluation Criteria and Metrics	Use of predictive and real-time indicators aligned with AI capabilities	Dashboard metrics, algorithmic indicators, and predictive evaluations	Hasselblad et al. (2024), He and Wang (2025), Anthony Jnr et al. (2022), Zhao et al. (2022), Sabbi et al. (2024)
Ethical and Legal Oversight	Ethical and legal structures that govern the use of AI in decision-making and service delivery	AI fairness, algorithm accountability, public oversight, and ethical auditing	López-López et al. (2018), Sabbi et al. (2024), He and Wang (2025), Chin and Guthrie (2023)
Institutional Capacity and Leadership	Leadership support, governance mechanisms, and resource capabilities that enable AI performance systems	IT staffing, leadership backing, knowledge systems, and organisational maturity	Heeks (2022), Almheiri et al. (2024), Waheduzzaman and Miah (2015)
Stakeholder Buy-In and Transition	Acceptance and participation of internal and external stakeholders in performance transitions involving AI	Communication workshops, codesign participation, and resistance mitigation	Chomchaiya and Esichaikul (2016), Chin and Guthrie (2023), Kenk and Haldma (2019), Anthony Jnr et al. (2022), Sabbi et al. (2024)

,

Table 2. Summary of performance management factors that address the research question.

Research Question	Performance Management Factor	Key Focus
Which performance management factors must be considered when using artificial intelligence in the local government sector?	Data Quality and Accessibility	Clean, interoperable, and timely data systems
	Strategic Alignment with Performance Goals	Aligning AI with goals and strategies
	Evaluation Criteria and Metrics	New, predictive, and real-time performance metrics
	Ethical and Legal Oversight	Accountability, fairness, and trust safeguards
	Institutional Capacity and Leadership	Human resources, skills, and governance capability
	Change Management and Stakeholder Engagement	Stakeholder inclusion and implementation transition

and

Table 3. Summary of performance management factors when using AI.

Performance Factor	Description	Relation to the Research Question	Key Studies
Data Quality and Accessibility	AI systems require clean, interoperable, and timely data to generate accurate and reliable performance insights.	AI’s contribution to performance depends on reliable and accessible data inputs.	Roztocki et al. (2023), Almheiri et al. (2024), Waheduzzaman and Miah (2015), Zhao et al. (2022), Anthony Jnr et al. (2022)
Strategic Alignment with Performance Goals	AI must be integrated into strategic planning cycles and performance frameworks, such as KPIs and Balanced Scorecards.	Aligning AI with local priorities ensures that it strengthens, rather than disrupts, strategic performance outcomes.	Karampotsis et al. (2024), Kenk and Haldma (2019)
Evaluation Criteria and Metrics	Performance evaluation must evolve to include real-time, predictive, and citizen-centric metrics enabled by AI.	AI-enabled monitoring transforms how performance is measured.	He and Wang (2025), Hasselblad et al. (2024), Chomchaiya and Esichaikul (2016), Zhao et al. (2022), Sabbi et al. (2024), Anthony Jnr et al. (2022)
Ethical and Legal Oversight	AI introduces ethical risks; performance systems must include transparency, equity, and accountability oversight.	AI tools must adhere to ethical and legal standards for legitimacy in performance evaluation.	López-López et al. (2018), Sabbi et al. (2024), He and Wang (2025), Chin and Guthrie (2023)
Institutional Capacity and Leadership	AI adoption requires skilled staff, adaptive leadership, and well-coordinated institutions to ensure successful implementation.	Without the internal capacity, AI cannot be meaningfully implemented for performance gains.	Heeks (2022), Roztocki et al. (2023), Almheiri et al. (2024), Waheduzzaman and Miah (2015)
Change Management and Stakeholder Engagement	AI impacts how services are delivered and managed; change processes and stakeholder engagement must be tracked and supported.	Performance systems must include engagement, training, and transition metrics to support AI-driven change.	Kenk and Haldma (2019), Chomchaiya and Esichaikul (2016), Sabbi et al. (2024), Chin and Guthrie (2023), Anthony Jnr et al. (2022)

detail the findings of the 22 peer-reviewed articles, which revealed that the successful application of AI in local government performance management depends on six critical performance management factors: namely, data quality and accessibility, strategic alignment with performance goals, evaluation criteria and metrics, ethical and legal oversight, institutional capacity and leadership, and change management and stakeholder engagement. These factors are interdependent and represent both technical and organisational dimensions of public administration. The comprehensive coding framework in the form of PM factors, definition, open code and contributing articles is presented in Table 1, followed by Table 2, which presents the summary of performance management factors that address the research question. Table 3 offers a summary of performance management factors when using AI.

To further understand which performance management aspects to consider when employing artificial intelligence in the local government sector, see Table 2 below for a summary of performance management factors that address the research question.

In addition to the thorough coding framework, it presents a summary of the performance management factors when using AI.

In conclusion, a total of six factors emerged, namely (1) data quality and accessibility, (2) strategic alignment with performance goals, (3) evaluation criteria and metrics, (4) ethical and legal oversight, (5) institutional capacity and leadership, and (6) change management and stakeholder engagement. As artificial intelligence (AI) becomes increasingly integrated into local governments’ governance and administrative systems around the world, it raises critical questions about how performance is managed, measured, and improved. AI is not a plug-and-play solution; its successful integration into local government operations hinges on several interdependent performance management factors, which were revealed through our rigorous synthesis of 22 peer-reviewed empirical studies.

4. Discussion

Evidence was limited to peer-reviewed empirical articles, excluding non-English and grey literature. This may have excluded innovative applications of AI occurring in practice but not yet published academically. Additionally, there was uneven geographic representation, with some regions contributing more heavily to the evidence base than others. While our data charting captured geographic contexts, the thematic synthesis inherent in a scoping review aimed to identify overarching trends and common performance management factors. Consequently, a detailed exploration of the nuances and differences in policy environments across various countries’ local governments and their specific implications for AI adoption was beyond the scope of this review and focused more on common trends relevant to these factors. This section gives a detailed discussion of the findings and management implications. While some of the identified factors might possess broader applicability, their detailed articulation and the associated challenges and opportunities are uniquely discussed and contextualised within the specific realm of local government performance management, drawing exclusively from studies meeting our stringent inclusion criteria for this sector. This contextualization helps highlight issues particular to AI adoption and impact in local governmental settings, differentiating it from other sectors like central government or corporate organisations.

4.1. Factor 1: Data Quality and Accessibility

One of the most foundational insights emerging from the literature is that high-quality, interoperable, and timely data is the foundation of AI-enabled performance management. Roztocki et al. (2023) and Almheiri et al. (2024) emphasised that without robust data governance, AI analytics will produce unreliable outputs. Waheduzzaman and Miah (2015) highlighted the role of e-government readiness and data maturity in ensuring that AI is effectively integrated. Anthony Jnr et al. (2022) showed that poor data systems in e-recruitment limited AI’s potential, while Zhao et al. (2022) demonstrated that credit risk prediction models in public–private partnerships require clean, timely datasets for accuracy. The identification of this factor directly answers the research question by demonstrating that AI’s ability to enhance performance measurement depends on data readiness. In environments where the data is siloed or inaccurate, AI outputs become misleading, thus compromising the integrity of performance management. Roztocki et al. (2023) and Almheiri et al. (2024) highlighted that public sector institutions require raw data and robust enterprise architectures to process and align their strategies with the performance targets. Similarly, Waheduzzaman and Miah (2015) developed a diagnostic framework that showed that e-government readiness, including data maturity, is a prerequisite for AI integration. Anthony Jnr et al. (2022) examined how poor data systems constrained the effectiveness of e-recruitment, while Zhao et al. (2022) illustrated that AI-based financial risk forecasting in PPPs depends on granular and timely datasets.

4.2. Factor 2: Strategic Alignment with Performance Goals

AI is not inherently strategic. For it to support public performance management, it must be aligned with the organisation’s overarching goals, strategic plans, and missions. Several studies caution against implementing AI as isolated pilot projects or experimental tools without long-term integration. The identification of this factor answers the research question by highlighting that AI should be used to reinforce, not replace, existing performance frameworks. Strategic alignment ensures that AI tools contribute to measurable outcomes that matter to stakeholders, from policy-makers to citizens. Karampotsis et al. (2024) showed how AI-enhanced Key Performance Indicators (KPIs) improved the alignment between strategic planning and operational execution in Greek municipalities. Kenk and Haldma (2019) reinforced the value of AI in performance alignment, particularly in the context of municipal reforms, where integration with long-term strategies allowed for more targeted and actionable strategies.

4.3. Factor 3: Evaluation Criteria and Metrics

A key finding from the review is that AI does not just change the speed or scale of performance monitoring—it transforms what aspects should be measured. Traditional input–output performance metrics often fail to capture the dynamic, predictive, and behavioural patterns that AI reveals. The identification of this factor answers the research question by emphasising that AI transforms what is measurable. It requires governments to rethink performance indicators and adopt multi-dimensional metrics that reflect predictive, preventative, and citizen-focused outcomes.

He and Wang (2025) exemplified this in municipal policing, where AI was used not just to count incidents but to predict and prevent crime. Hasselblad et al. (2024) found that the use of AI in social services required new evaluation frameworks to align predictive insights with human-centred outcomes. Chomchaiya and Esichaikul (2016) integrated AI into procurement monitoring; Zhao et al. (2022) used machine learning for financial risk assessment; Sabbi et al. (2024) highlighted institutional trust as a performance metric; and Anthony Jnr et al. (2022) illustrated AI’s role in real-time recruitment analytics.

4.4. Factor 4: Ethical and Legal Oversight

Integrating AI into local government necessitates rethinking the ethical standards and legal safeguards embedded in performance management systems. When AI influences eligibility decisions, predictive policing, or resource allocation, it can introduce automation bias, reduce transparency, or erode public trust. When applied without ethical safeguards, AI can undermine the legitimacy of performance management. This factor thus relates to the research question in that it shows that performance should include service outputs and procedural justice, equity, and democratic accountability. López-López et al. (2018) showed that reputational performance in municipalities improved when transparency protocols were embedded into AI-powered e-government portals. Sabbi et al. (2024) stressed that accountability must be built into AI systems, while He and Wang (2025) warned of automation bias in policing. Chin and Guthrie (2023) highlighted the governance tensions in AI-driven mobility projects, showing how opaque algorithms can challenge political accountability. The importance of trust, transparency, and fairness was also emphasised in Sabbi et al. (2024), who concluded that AI must operate under a performance logic that extends beyond numerical efficiency and encompasses procedural justice.

4.5. Factor 5: Institutional Capacity and Leadership

While the technical potential of AI is often emphasised, the literature is equally clear that its successful integration into performance systems depends on institutional capacity and leadership. Local governments must possess the digital infrastructure and leadership required to interpret AI insights, foster collaboration, and manage risk. This factor addresses the organisational readiness required for AI to contribute meaningfully to performance. Without human and institutional capacity, AI becomes a symbolic gesture rather than a transformational tool. Thus, it is a core consideration when planning AI-enabled performance initiatives. Heeks (2022) outlined how AI-enabled performance initiatives in Ghana struggled due to institutional bottlenecks, including underinvestment in skills and leadership. Roztocki et al. (2023) and Almheiri et al. (2024) emphasised the role of enterprise architecture and dynamic capabilities in building readiness. Waheduzzaman and Miah (2015) affirmed that without strong governance structures and internal systems, AI efforts flounder regardless of the technology’s potential.

4.6. Factor 6: Change Management and Stakeholder Engagement

AI does not exist in a vacuum—it alters workflows, decision-making hierarchies, and service logic. Successful performance management in AI-enabled local government requires intentionally focusing on organisational transition and stakeholder engagement. AI entails organisational and cultural change. The identification of this factor underscores the importance of managing perceptions, expectations, and power dynamics in driving performance transformation. Stakeholder inclusion must be embedded in performance metrics—not just for planning, but also for monitoring and evaluation. Kenk and Haldma (2019) illustrated that municipal reforms using AI benefited from structured change management strategies. Chomchaiya and Esichaikul (2016) embedded stakeholder satisfaction into AI procurement frameworks. Anthony Jnr et al. (2022) and Sabbi et al. (2024) showed that participatory approaches increase buy-in, while Chin and Guthrie (2023) warned that excluding stakeholders from algorithmic decision-making can erode legitimacy.

In conclusion, the synthesis of the 22 peer-reviewed studies shows that the effective use of artificial intelligence in local government performance management is shaped by six interdependent performance management factors: data quality and accessibility, strategic alignment with performance goals, evaluation criteria and metrics, ethical and legal oversight, institutional capacity and leadership, and change management with stakeholder engagement. Each factor reflects both technical and organisational considerations, underscoring that AI’s contribution to performance improvement is not solely determined by algorithmic capability, but by the readiness and adaptability of the performance management environment. Data quality and accessibility emerged as the foundational requirement, with studies consistently showing that without reliable, timely, and interoperable datasets, AI systems produce inaccurate or misleading outputs (Roztocki et al., 2023; Almheiri et al., 2024; Waheduzzaman & Miah, 2015; Anthony Jnr et al., 2022; Zhao et al., 2022). Strategic alignment ensures that AI is embedded into broader planning and monitoring frameworks, preventing fragmented initiatives and ensuring that the outputs support long-term performance goals (Karampotsis et al., 2024; Kenk & Haldma, 2019).

Evaluation criteria and metrics must evolve to incorporate predictive, citizen-focused, and real-time indicators that capture AI’s added value beyond traditional input–output measures (He & Wang, 2025; Hasselblad et al., 2024; Chomchaiya & Esichaikul, 2016; Zhao et al., 2022; Sabbi et al., 2024; Anthony Jnr et al., 2022). Ethical and legal oversight is critical to maintaining legitimacy, fairness, and transparency in AI-enabled decision-making, particularly in sensitive service areas such as policing, mobility, and public service eligibility (López-López et al., 2018; Sabbi et al., 2024; He & Wang, 2025; Chin & Guthrie, 2023). Institutional capacity and leadership directly influence whether AI adoption translates into measurable performance gains, with leadership, skills, and governance structures proving decisive (Heeks, 2022; Roztocki et al., 2023; Almheiri et al., 2024; Waheduzzaman & Miah, 2015).

Finally, change management and stakeholder engagement ensure that AI implementation is accepted, understood, and supported by those who must use it or are impacted by it, reinforcing trust and adoption (Kenk & Haldma, 2019; Chomchaiya & Esichaikul, 2016; Anthony Jnr et al., 2022; Sabbi et al., 2024; Chin & Guthrie, 2023). Collectively, these six factors highlight that AI is not a plug-and-play performance tool. Its value in the local government sector depends on institutional readiness and embedding it within robust data systems, aligning it with strategic objectives, adapting performance metrics, ensuring adherence to ethical standards, and actively engaging stakeholders throughout the change process.

5. Conclusions

Although the review protocol was not registered, all the procedures regarding article selection, inclusion and exclusion criteria, and data synthesis followed the principles of PRISMA. The protocol was not registered due to the nature of the study. This research sought to determine the performance management factors that must be considered when using artificial intelligence in the local government sector. The findings indicate that six performance management factors—namely, data, strategy, metrics, ethics, capacity, and change—form a cohesive framework for understanding what must be in place for AI to deliver value in local government performance systems. The findings of this study show that these six factors are interconnected and are required for AI to be deployed successfully and provide useful results. In practice, AI within the global local government won’t offer meaningful outcomes unless these six factors are addressed. Therefore, the findings of the study form a comprehensive guide for policymakers who intend to integrate AI successfully within the local government. This study highlights that AI entails more than innovation; it reshapes the foundations of performance governance, requiring new capabilities, values, and institutional practices. The findings of this study provide major conclusions in the body of knowledge. This study offers critical findings on how the integration of AI will provide measurable value.

One of the limitations of this study is its reliance on secondary data, and thus, it does not include new empirical exploration through in-depth case studies focusing on a specific type of local government. Although the study used five reputable databases, these databases do not cover the entire body of literature. However, other metrics, such as institutional distribution, current study, and temporal distribution, were selected to capture trends over time, which aligns with our objective of understanding longitudinal shifts. The institutional distribution was excluded because of its inconsistency in data affiliation across data sources, which could compromise reliability. Gaps in the selected databases could be insufficient information (too few studies or scarce data), at risk of bias (methodologically flawed studies), inconsistent results (conflicting findings among studies), or not the right information (studies not applicable to the review’s population or setting). A comprehensive framework could have been used to characterise these gaps by applying the Patient, Intervention, Comparison, and Outcome (PICO) framework to describe the knowledge deficiency and categorising the reasons for the gaps. Constraints of the PRISMA protocol include inconsistent and inadequate author adherence and insufficient verification by editors and peer reviewers, leading to incomplete reporting even when authors claim to follow PRISMA. Other constraints involved the guidelines not assessing quality or the conduct of a review, the potential for over-reliance leading to exclusion of valid studies, the existence of multiple PRISMA versions (for example, PRISMA 2009 vs. PRISMA 2020), and the lack of equivalent “active” technology-driven implementation strategies compared to other reporting guidelines. The inclusion of only English publications introduced a language bias, which potentially excluded relevant research, skewing conclusions and limiting the diversity of perspectives in the review. This exclusion could have led to an incomplete understanding of global performance management and may have significantly impacted the generalisability and reliability of the review’s findings. The last limitation of the study is that data collection was limited to just two months, which may not adequately represent the entire year of 2025.

This systematic mapping of the literature, however, provides a crucial foundation for future research, which we recommend should include such empirical studies to explore specific real-world situations and the influence of AI on employee performance within local governments in greater depth. The study focused on studies conducted on local governments around the world; therefore, it is recommended to compare government and private sector institutions. Future studies should be conducted on how the use of AI in PM influences employees’ performance within local governments.