Optimization of organizational processes in healthcare using artificial intelligence and implications for BRICS countries: a systematic review with narrative synthesis

Introduction

Modern healthcare systems face increasing challenges associated with population ageing, the growing prevalence of chronic diseases, shortages of healthcare personnel and financial resources, an expanding administrative workload, and the need to improve operational efficiency. In this context, artificial intelligence (AI) technologies are increasingly recognized not only as tools for clinical decision support but also as promising solutions for optimizing organizational and managerial processes in healthcare [1–4]. While research conducted before 2020 primarily focused on the diagnostic performance of AI algorithms, recent studies have shifted toward integrating AI into organizational workflows and evaluating its impact on healthcare performance indicators. Davenport et al. and Bhagat et al. suggested that the greatest organizational and economic benefits of AI in healthcare may arise from the optimization of administrative processes, logistics, and patient engagement [1][5].

At the current stage of digital transformation, AI is increasingly integrated into enterprise resource planning systems, electronic health records, data analytics platforms, and patient relationship management systems, enabling automation of routine administrative tasks, more efficient resource allocation, and data-driven managerial decision-making. The rapid emergence of generative AI has further accelerated this transformation by expanding opportunities for automated clinical documentation, natural language processing, and patient communication. Despite the rapid growth of this field, relatively few systematic reviews have comprehensively examined the organizational applications of AI in healthcare, particularly with regard to implementation effectiveness, economic outcomes, organizational barriers, and strategies for successful adoption [1][6–9].

The implementation of AI is of particular importance for BRICS countries, whose healthcare systems face increasing service demand, workforce shortages, and substantial heterogeneity in digital maturity. Available evidence, derived predominantly from the Russian Federation, India, and Brazil, suggests that AI has considerable potential to improve organizational efficiency, optimize resource utilization, and strengthen healthcare management under resource-constrained conditions [10–13]. However, evidence from the remaining BRICS member states remains limited, highlighting the need for further research and a cautious interpretation of the current findings.

Therefore, the aim of this systematic review with narrative synthesis was to identify, critically evaluate, and synthesize the available evidence on the impact of AI technologies on organizational processes in healthcare and to assess the potential applicability of these approaches to selected healthcare systems in BRICS countries.

Methods

Study design and reporting framework

This systematic review with narrative synthesis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses, PRISMA 2020. The review protocol was developed by the authors prior to study selection and predefined the eligibility criteria, literature search strategy, data extraction process, quality assessment, and methods for evidence synthesis. The review protocol was not registered in the PROSPERO database because of the review’s primary focus on organizational and managerial aspects of AI implementation in healthcare and the substantial methodological heterogeneity of the included studies.

Databases and search strategy

The literature search was conducted between January and February 2026 using PubMed and the major Russian scientific databases, including eLibrary.ru (incorporating the Russian Science Citation Index) and CyberLeninka. Search strategies (Table) were adapted for each database with no language restrictions applied. Peer-reviewed publications were prioritized while selected preprints were used only as contextual evidence.

Search strategies were adapted to the syntax and indexing systems of individual databases. The searches covered publications from January 2018 to December 2025. Reference lists of relevant reviews and eligible studies were additionally screened to identify potentially missed publications.

Study selection

A total of 524 records were identified, including 268 from PubMed, 118 from eLibrary.ru (including publications indexed in the Russian Science Citation Index), and 138 from CyberLeninka. Duplicate records were removed manually. After removal of duplicate records (n = 48), 476 records remained for title and abstract screening. Title and abstract screening was performed independently by two authors (Ryazanov A.S. and Makarovskaya M.V.). Disagreements were resolved through discussion, and, when necessary, the opinion of a third author (Cherepov A.V.) was sought. Of these, 356 publications were excluded because they did not meet the eligibility criteria. The full texts of 120 articles were assessed for eligibility, and 42 records were excluded following full-text review. Consequently, 78 publications met the inclusion criteria and were included in the final qualitative synthesis. The study selection process is presented in Figure.

Eligibility criteria

The inclusion criteria were as follows: (1) publications published between 2017 and 2026; (2) primary empirical studies, implementation studies, observational studies, systematic reviews, and narrative reviews, methodological papers, governance-related publications, and other contextual sources addressing organizational or managerial applications of AI in healthcare; and (3) publications reporting organizational, operational, economic, or implementation-related outcomes or addressing methodological and governance issues relevant to AI implementation. Primary empirical and implementation studies were considered the core evidence base, whereas reviews, methodological papers, governance-related publications, and preprints were used to support contextual interpretation.

Full-text articles were excluded if they: (1) did not address organizational, operational, managerial, economic, or implementation-related aspects of AI in healthcare; (2) focused exclusively on clinical diagnostic or therapeutic performance without organizational or managerial implications; (3) lacked sufficient methodological information; (4) were editorials, commentaries, conference abstracts, or opinion papers without substantive organizational, methodological, governance, or implementation-related content; or (5) were not available in full text.

Data extraction

The following data were extracted from each included study: first author, year of publication, country, study design, healthcare setting (e.g., emergency department, inpatient care, or administrative environment), type of AI technology, targeted organizational or managerial process (such as resource planning, clinical documentation, logistics, or financial administration), and the main organizational outcomes and performance indicators. Data extraction was performed using a standardized data extraction form to ensure consistency and accuracy across all included studies.

Risk of bias and quality assessment

Given the methodological heterogeneity and the inclusion of both primary and secondary publications, a structured descriptive methodological appraisal rather than a formal quantitative risk-of-bias assessment was performed. The appraisal considered the clarity of study objectives, transparency of methods, description of the sample and setting, comparator selection, completeness of outcome reporting, and the distinction between model performance and real-world organizational outcomes. Identified methodological limitations were considered during the narrative synthesis and interpretation of the findings.

Data synthesis

A quantitative meta-analysis was not performed because of substantial methodological and clinical heterogeneity among the included studies. The reviewed publications differed considerably in study design, healthcare settings, patient populations, organizational contexts, AI technologies, implementation approaches, outcome measures, and reporting methods. The studies evaluated diverse domains, including patient flow management, emergency department triage, operating room scheduling, documentation support, revenue cycle management, logistics, supply chain optimization, and patient communication systems. As a result, pooling effect estimates into a single quantitative measure would not have been methodologically appropriate.

Therefore, a narrative synthesis approach was adopted. Accordingly, the present study should be regarded as a systematic review with narrative synthesis rather than a conventional systematic review restricted exclusively to primary empirical studies. This approach was considered appropriate because of the methodological heterogeneity of the available literature and the objective of synthesizing evidence across multiple organizational domains. The included studies were grouped into predefined thematic categories reflecting major organizational applications of AI in healthcare. Findings were synthesized descriptively, with emphasis on implementation characteristics, organizational outcomes, operational efficiency indicators, and relevance to healthcare systems in BRICS countries and other resource-constrained settings.

Results

Characteristics of included publications

A total of 78 records met the inclusion criteria. The included publications demonstrated substantial methodological heterogeneity and comprised systematic reviews, narrative reviews, implementation studies, observational studies, cross-sectional surveys, modeling studies, algorithm development studies, randomized clinical trials, and healthcare management case studies. Most studies originated from the United States, Europe, and international collaborative groups, whereas several studies were conducted in BRICS countries, including Russia, India, and Brazil.

Study duration, sample sizes, patient-level demographic characteristics, including sex distribution, control groups or comparators was reported inconsistently and therefore could not be summarized quantitatively. In most implementation and observational studies, outcomes were assessed using historical comparisons, workflow performance metrics, before–after analyses, or operational benchmarks. Detailed characteristics of all 78 included publications are provided in Supplementary Tables 1 and 2 (supplementary materials on the journal website https://doi.org/10.47093/3034-4700.2026.3.1.31-49-annex).

Organizational applications of artificial intelligence in healthcare

The included publications covered a wide range of organizational and managerial uses of AI in healthcare, from healthcare operations and automation of administrative tasks to healthcare delivery, practical AI implementation, and management-focused digital transformation [1–4][9][11]. The largest group of included studies evaluated predictive models for emergency departments and inpatient care. These AI solutions demonstrated high predictive performance in selected studies (Area Under Curve [AUC] up to 0.92–0.97). Their implementation was associated with reductions in patient waiting time of 18–26% and more efficient resource allocation within healthcare organizations [14][15]. Other studies and reviews investigated the use of machine-learning methods to predict patient flow, support emergency department triage, manage overcrowding, estimate the probability of hospital admission and waiting times, and improve patient throughput [16–25].

Another major category comprised AI applications for operating room management. Machine learning algorithms capable of predicting surgical procedure duration and postoperative events, procedural scheduling, operating room usage-time estimation, and operating room efficiency optimization [26–34]. These systems were associated with reduction in operating room idle time and improved scheduling accuracy (F1-score up to 0.78–0.82) in selected implementation settings [26].

Studies investigating AI applications in healthcare logistics and supply chain management demonstrated improved demand forecasting, optimization of inventory levels, and reduced risk of shortages of critical medical supplies. Machine learning, neural networks, synthetic data and generative AI approaches supported pharmaceutical and medical device demand forecasting [35–39]. More recent evidence further supports the role of AI in pharmaceutical demand forecasting, inventory optimization, and supply chain resilience [38][39].

Process mining techniques are used to identify process bottlenecks, optimize patient pathways, and improve operational efficiency. For example, one study reported about reduction in delays in care delivery following the identification and elimination of hidden organizational inefficiencies through process mining analysis [40].

AI-assisted clinical documentation, ambient documentation, natural language and generative AI applications facilitated automated generation of clinical documentation, improved documentation completeness, and were associated with lower levels of physician administrative burden and burnout, with burnout prevalence decreasing in individual reports. Further studies and reviews of generative AI, ambient documentation systems, and AI-assisted note preparation reported a lower administrative burden, more efficient documentation, and several workflow improvements in selected clinical settings [7][41–46]. At the same time, conversational AI systems, patient-oriented chatbots, voice technologies, and remote digital tools were used to facilitate communication with patients, automate responses to routine questions, support chronic disease management, and improve access to healthcare information [47–53].

In revenue cycle management (RCM), AI was applied to automate insurance claims processing and detect fraudulent or anomalous transactions; it improved the efficiency of claims processing, reduced the likelihood of administrative errors, and facilitated the identification of fraudulent or anomalous transactions [54–57]. Additional publications considered the application of AI to medical billing and coding, care management, health insurance processes, and the detection of fraudulent medical claims. Some studies specifically examined explainable graph-based methods for identifying fraud [58][55–57].

Economic, organizational, and implementation-related outcomes

Although economic outcomes were rarely evaluated as primary endpoints, the included studies reported short-term organizational and economic benefits associated with healthcare operations, administrative automation, documentation support, patient communication, logistics, claims processing, and resource planning [1–4][7][9][11][37–39][46][47][50–58].

AI applications in RCM reduced insurance claims processing time, decreased coding errors and document reprocessing, and contributed to lower administrative costs for healthcare organizations [54–58]. AI-based patient flow management and demand forecasting systems reduced patient waiting times by 18–26%, improved patient distribution across clinical units, and enhanced the utilization of hospital beds and staff resources [15][16][18][19][24]. Similarly, studies evaluating AI-assisted operating room scheduling reported more accurate predictions of surgical case duration and more efficient use of high-cost surgical resources and lower indirect operational costs [26–34].

AI-assisted clinical documentation systems and generative AI technologies reduced documentation with administrative workload, improved workforce productivity, and enabled clinicians to devote more time to direct patient care [7][41–46]. Likewise, conversational AI systems capable of automating routine patient inquiries reduced the workload of administrative personnel while maintaining access to healthcare services. Patient-facing conversational agents improved healthcare accessibility, facilitated patient communication, and supported administrative efficiency across different healthcare settings [47][48][50–53].

Despite evidence of potential operational costs reductions, only a limited number of studies performed comprehensive economic evaluations using indicators such as return on investment (ROI), payback period, or total cost of ownership (TCO). Most studies focused on improving healthcare organization, workflow efficiency, and resource use rather than assessing formal economic outcomes. Consequently, current evidence primarily reflects short-term organizational and economic outcomes, whereas the long-term economic effectiveness and sustainability remain insufficiently studied [59–61].

Implementation-related publications reported that AI adoption was influenced by data infrastructure, interoperability, workflow redesign, organizational readiness, digital transformation capacity, interdisciplinary collaboration and governance mechanisms for monitoring AI performance after deployment [59][62–70]. Other sources emphasized that bias, fairness, explainability, cybersecurity, ethics, regulation, and responsible AI governance were frequently discussed as implementation-related risks rather than as directly measured organizational outcomes [8][71–78].

Overall, AI implementation was associated with improvements in operational performance indicators, including reduced patient waiting times, decreased operating room idle time, more efficient clinical documentation, enhanced administrative processes, improved resource utilization, reduced administrative burden, and workflows optimization.

Evidence from BRICS countries

Although most of the available evidence originated from high-income countries, published studies from the Russian Federation, India, and Brazil suggest that similar AI-based organizational solutions can be successfully adapted to selected healthcare systems within BRICS countries. However, evidence from the remaining BRICS member states remains limited, underscoring the need for further research to evaluate the effectiveness, scalability, and sustainability of AI implementation across diverse healthcare settings [6][10][35][40][49][61][79][80].

Parven to studies conducted in BRICS countries, as they demonstrate the feasibility of adapting AI-based organizational solutions to healthcare systems operating under resource constraints and increasing service demand. In India, machine learning models applied to pharmaceutical demand forecasting and supply chain management improved forecasting accuracy and reduced logistical losses [35]. Additional India/global evidence addressed remote monitoring and workflow-related AI applications relevant to healthcare delivery under resource-constrained conditions [49]. In Brazil, a review showed that process mining techniques were successfully used to analyze patient pathways and identify hidden organizational inefficiencies, resulting in reduced delays in care delivery and greater transparency of healthcare processes [40].

Several studies have also reported favorable results for AI-supported patient flow management and hospital capacity forecasting, particularly in large multidisciplinary healthcare centers [4][11][16]. In publications also described digital health development, information technologies for improving the organization of medical care, and the national implementation of AI technologies in healthcare [6][61][79].

Overall, the available evidence from BRICS countries remains limited but indicates that organizational AI applications can be successfully adapted to diverse healthcare environments when supported by adequate digital infrastructure, interoperable information systems, and appropriate governance frameworks.

Artificial intelligence implementation in the Russian healthcare system

Evidence from the Russian Federation also demonstrates the growing role of AI technologies in healthcare organization and management. One of the largest initiatives is the Moscow Experiment on Computer Vision Technologies coordinated by the Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Healthcare Department. During the first year of implementation, 18 AI systems were integrated into the Unified Radiological Information Service and became available to 538 radiologists, demonstrating the feasibility of large-scale AI deployment in routine clinical practice [81].

Subsequent study involving more than 1.3 million imaging examinations confirmed the feasibility of integrating AI services into routine radiology workflows and highlighted the need for continuous performance monitoring and quality control of diagnostic processes [66]. Another study reported a reduction in computed tomography scan interpreting and reporting time in an inpatient COVID-19 setting, indicating the potential organizational relevance of AI-assisted radiology workflows in high-demand clinical environments [80].

In addition, the Unified Medical Information and Analytical System has provided the digital infrastructure necessary for AI-assisted patient routing, appointment scheduling, resource management, and administrative decision support. These developments provide example of large-scale AI implementation in a publicly funded healthcare environment [79][80]. Together with publications on Russian digital health development and information technologies for medical care organization, these publications indicate that the Russian evidence base includes both large-scale implementation experience and contextual sources on healthcare digital transformation [6][61][79].

Discussion

Principal findings

This systematic review with narrative synthesis demonstrates that AI has considerable potential to improve the organizational performance of healthcare systems by optimizing patient flow management, resource utilization, clinical documentation, and administrative processes. The strongest evidence was identified for predictive analytics, AI-assisted clinical documentation, and operational management applications, all of which were consistently associated with improvements in healthcare efficiency and organizational performance. The findings also suggest a shift in the role of AI from automation of individual tasks integration into organizational workflows and managerial decision-making. This shift indicates that AI should be considered not only as a technological innovation but also as a managerial instrument requiring organizational readiness, institutional capacity, and effective governance.

The principal contribution of this systematic review lies in its comprehensive synthesis of evidence regarding the organizational and managerial applications of AI in healthcare rather than its clinical applications. Unlike most previous reviews, which primarily focus on algorithm performance and clinical outcomes, the present study integrates organizational, operational, and economic dimensions of AI implementation within a unified analytical framework. Furthermore, it proposes a management-oriented classification of AI applications according to their primary organizational function, including emergency department management, operating room optimization, revenue cycle management, logistics, and clinical documentation. Finally, the review highlights the potential transferability of these approaches to selected BRICS healthcare systems while identifying key institutional, organizational, and governance-related challenges that may influence successful implementation.

Implementation barriers

The available evidence indicates that the barriers to successful AI implementation include technological, organizational, managerial, and ethical rather than purely technological. Recent evidence suggests that technological immaturity (77%) and financial constraints (47%) represent more significant obstacles to AI implementation than clinician resistance (17%) or insufficient managerial support (7%) [68][73]. Thus, successful AI implementation depends not only on technology itself but also on organizational readiness, institutional capacity, workforce competencies, and effective governance.

The effectiveness of AI-based solutions is largely determined by the quality of the data used for model development and implementation. Fragmentation information systems, lack of unified coding standards, and unstructured electronic health records data limit reliable AI use, as illustrated by datasets such as MIMIC-CXR. Addressing this “digital chaos” requires centralized data infrastructure, unified data models, robust data governance, and standardized interoperable data [62–65].

AI implementation also transforms clinical and administrative workflows and may raise concerns about job security, professional autonomy, clinical responsibility, and performance monitoring. The “AI as a second reader” model may promote clinician acceptance by preserving final clinical responsibility while providing decision support [59][67][70][71].

The safe and effective integration of AI systems requires not only technical solutions, including Fast Healthcare Interoperability Resources- and Observational Medical Outcomes Partnership-based interoperability frameworks, but also new regulatory approaches [82]. For example, the Food and Drug Administration (FDA) has developed the concept of Predetermined Change Control Plans (PCCPs) for AI-enabled medical devices, aiming to balance innovation with patient safety and regulatory oversight [64][81][83].

Ethical and regulatory challenges represent some of the most important barriers to the large-scale implementation of AI in healthcare. One of the major concerns is algorithmic bias, whereby AI models trained on historical datasets may inherit and amplify existing healthcare inequalities. The landmark study by Obermeyer et al. demonstrated racial bias in an algorithm used to allocate healthcare resources, which systematically underestimated the healthcare needs of Black patients [74]. More recent studies confirm that the underrepresentation of minority populations and dominated Western, Educated, Industrialized, Rich, and Democratic (WEIRD) population datasets, may limit AI model generalizability and increase the risk of biased decision-making highlighting the need for continuous fairness assessment and algorithm auditing [8][9][75][84].

Another major challenge concerns the transparency of AI-driven decision-making, often referred to as the “black-box” problem. The limited interpretability may reduce clinicians’ trust and hinder regulatory acceptance. Therefore, explainable AI is important for regulatory compliance, human oversight and high-risk healthcare applications, in line with requirements for transparency, human oversight, and risk management [9][71][72][76][77].

Cybersecurity represents another critical challenge. The integration of AI into healthcare information systems expands the potential attack surface for cyber threats. In particular, adversarial attacks, which intentionally manipulate input data to alter AI predictions, pose emerging risks to the reliability and safety of AI-assisted healthcare services. At present, the global regulatory landscape remains fragmented, and further international harmonization of cybersecurity standards and AI governance frameworks is required to ensure the safe and responsible deployment of AI technologies in healthcare [78][85][86].

Economic sustainability and practical implementation

Economic sustainability should also be considered when evaluating AI implementation. Many reported organizational benefits do not fully account for the costs of system integration, cybersecurity, staff training, workflow redesign, and long-term maintenance. Available evidence suggests that enterprise AI projects in healthcare may achieve a median ROI of approximately 3.2:1, with an estimated payback period of 12–18 months. However, comprehensive economic evaluations remain limited. Therefore, TCO analyses should account for additional implementation costs, including system integration, cybersecurity, workforce training, and continuous monitoring of financial performance [59][60].

From a practical perspective, the successful implementation of AI requires strategic planning, phased deployment, workforce development, data interoperability, and effective governance. AI initiatives should be aligned with organizational priorities and supported by senior leadership, while pilot testing, evaluation, refinement, and gradual scaling, should be accompanied by continuous monitoring of operational, clinical, and ethical performance indicators [59][69].

Successful implementation depends on multidisciplinary collaboration involving clinicians, healthcare managers, data scientists, engineers, and ethics experts, as well as a data-driven organizational culture and digital competencies among healthcare professionals [87]. Contemporary implementation frameworks, including the IA²TF model, emphasize the integration of data interoperability, continuous performance monitoring, ethical oversight, and adaptive governance mechanisms as essential components of sustainable AI adoption [8][59][69][70].

As AI applications mature and expand, healthcare organizations should establish comprehensive governance frameworks covering the entire AI lifecycle. Such frameworks should be aligned with external regulatory requirements, including the European Union Artificial Intelligence Act and the FDA Software as a Medical Device guidance, and should incorporate regular algorithm audits, monitoring of model drift, risk management, and continuous quality assurance [8][27][59][88][89].

Finally, implementation strategies should be guided by measurable organizational and economic outcomes. Healthcare organizations should define clear operational and financial performance indicators before implementation and conduct structured post-implementation evaluations, including assessments of ROI, TCO, implementation costs, and workforce training requirements. Together, these principles may facilitate the transition from isolated pilot projects to sustainable organizational integration of AI technologies [9][59][61][68].

Implications for BRICS countries

Although most of the included studies were conducted in high-income countries, the findings of this review have important implications for healthcare systems in BRICS countries [10][11][72]. For selected BRICS countries, particularly the Russian Federation, India, and Brazil, the available evidence suggests that scalable and economically feasible AI solutions may support organizational improvements under resource-constrained conditions. However, successful implementation depends not only on technological innovation but also on digital infrastructure, data interoperability, governance frameworks, workforce competencies, and context-specific implementation strategies.

In this context, structural limitations are especially important. Healthcare systems across BRICS countries differ considerably in terms of digital maturity, the availability of interoperable health information systems, and the overall quality of data infrastructure [63][64][71][72]. Fragmented information systems, partial digitalization, incomplete or non-standardized healthcare data, and limited interoperability may reduce the scalability and generalizability of predictive and optimization models [62–64][90]. Governance-related issues also remain important, including standards for data governance, algorithm validation, accountability, post-deployment monitoring, cybersecurity, and regulatory oversight [8][71][72][76][78].

Financial limitations may create additional difficulties for AI implementation in resource-constrained healthcare systems. In such settings, immediate operational needs can reduce the ability of healthcare organizations to invest in digital infrastructure, system integration, cybersecurity, and staff training over the long term [59][60][69]. At the same time, evidence from India and Brazil suggests that AI can still provide practical value when it is applied to clearly defined operational problems, such as pharmaceutical demand forecasting, healthcare logistics, patient pathway analysis, and administrative efficiency [16][35][40]. These findings suggest that successful implementation depends on alignment with local organizational priorities, infrastructure capacity, governance mechanisms, and healthcare needs rather than on technological sophistication alone.

Accordingly, AI implementation in BRICS countries should follow a phased, context-sensitive approach [10][59][69][70]. Priority should be given to AI applications that combine relatively low implementation costs with measurable organizational benefits, particularly in operational planning, logistics, clinical documentation, and patient flow management. Gradual integration with existing health information systems, continued investment in digital infrastructure and data standards, and the development of digital and analytical competencies among healthcare professionals should constitute key components of national implementation strategies [8][59][69–71]. The Russian Federation, India, and Brazil, may serve both as settings for adapting AI technologies developed in high-income countries and as contributors to context-specific organizational and managerial AI solutions in healthcare. However, further studies from other BRICS member states are required to establish scalable, economically sustainable, and context-specific models for AI implementation across the broader BRICS region.

Limitations

This systematic review with narrative synthesis has several limitations. Although studies from Russia, India, and Brazil were included, most evidence originated from high-income countries, limiting transferability to all BRICS healthcare systems. Substantial heterogeneity in study design, settings, AI technologies, outcomes, and implementation strategies precluded meta-analysis and limited direct comparisons. Few studies reported comprehensive economic evaluations, including ROI, TCO, or long-term cost-effectiveness; therefore, current evidence mainly reflects short-term organizational and operational outcomes. The evidence base included empirical studies, implementation studies, and secondary reviews, limiting direct causal inference. Potential publication bias and the predominance of studies reporting favorable implementation outcomes may also have influenced the synthesis. Finally, evidence from resource-constrained healthcare systems remains limited, particularly for several BRICS member states. Unresolved questions concern long-term economic effectiveness, scalability, sustainability, routine workflow integration, and applicability to resource-constrained healthcare systems.

Future directions

The available evidence suggests that the next stage of AI development in healthcare is likely to involve a transition from isolated automation tools toward integrated organizational decision-support systems. Generative AI may support automated clinical documentation, analytical reporting, and management-oriented decision support, thereby reducing administrative workload and improving organizational efficiency.

Digital twins, federated learning, and privacy-preserving computing represent further promising directions, as they may support simulation of patient flow, resource utilization, and organizational changes and collaborative AI models development without transferring sensitive patient-level data. These technologies may be particularly relevant for BRICS healthcare systems, where data interoperability, information security, and heterogeneous digital infrastructure continue to represent important barriers to large-scale AI implementation.

Future progress in healthcare AI will depend not only on continued advances in machine learning algorithms but also on the ability of healthcare systems to integrate AI technologies into existing organizational, managerial, and regulatory frameworks. Sustainable implementation will require coordinated development of digital infrastructure, governance mechanisms, workforce competencies, and evaluation of organizational and economic impact.

Conclusion

This systematic review with narrative synthesis shows that AI can improve several organizational and managerial processes in healthcare, especially patient flow management, resource allocation, operating room scheduling, clinical documentation, and administrative workflows. At the same time, the available evidence remains heterogeneous and mostly describes short-term organizational effects. Long-term economic effectiveness, scalability, sustainability, and transferability of these technologies, particularly within BRICS healthcare systems, are still not sufficiently studied. In general, AI should be viewed not only as a clinical technology, but also as an important tool for healthcare management and organizational transformation. Its sustainable implementation will depend on successful integration into existing workflows, as well as on adequate digital infrastructure, trained personnel, and effective governance.