Healthcare supply chains generate heterogeneous multimodal data - medical images, biosensor readings, clinical text, and transactional records - that are currently processed in isolation, resulting in fragmented decision-making and suboptimal sustainability outcomes. Key bottlenecks include: (a) data integrity risks from centralised storage systems, (b) lack of real-time cross-modal visibility across supply chain stages, (c) consensus inefficiency in distributed ledger systems leading to high computational overhead, and (d) absence of integrated AI reasoning across data modalities. A multimodal approach is necessary because single-modal systems cannot capture the complex interdependencies among these data streams. This study develops a Multimodal Intelligent Hyperautomation framework by integrating Blockchain Technology (BCT), the Internet of Medical Things (IoMT), and Artificial Intelligence (AI). The framework supports cross-modal data fusion and facilitates adaptive decision-making across healthcare supply chain processes. Data were collected from 316 respondents in 210 Chinese hospitals through a structured survey. (H1) BCT utilisation positively impacts IoMT utilisation; (H2) BCT utilisation positively impacts AI utilisation; (H3) IoMT utilisation positively impacts AI utilisation; (H4) IoMT utilisation positively impacts healthcare supply chain sustainability; and (H5) AI utilisation positively impacts healthcare supply chain sustainability. Analysis via Structural Equation Modelling with data from 316 respondents across 210 Chinese hospitals confirmed all five hypotheses (all p < 0.001). Drawing on these empirical results, researchers designed a multimodal fusion layer to integrate visual, textual, and sensor data streams within a blockchain-secured hyperautomation pipeline. Researchers also introduce an Improved Byzantine Fault Tolerance (IMBFT) consensus mechanism that achieves higher throughput and greater fault tolerance than conventional methods. In addition, the proposed cross-modal architecture promotes explainable and human-centric automated processes. This study extends current understanding of multimodal AI applications in intelligent hyperautomation and offers actionable technical designs to enhance sustainability in healthcare supply chains. Unlike prior studies that examine individual technologies in isolation, this study contributes the first integrated multimodal hyperautomation framework for healthcare supply chains, yielding measurable benefits: 20-25% reduction in hospital drug inventory, 23% higher consensus throughput via the IMBFT algorithm, and enhanced cross-modal predictive accuracy (R-2 = 0.946) for demand forecasting compared to SVM, BP, ARIMA, and hybrid baseline models.
Small and medium enterprises (SMEs) are increasingly relying on digital technology to enhance production efficiency. For SMEs in the manufacturing industry, orders typically include the buyer's input information. As SMEs lack the capacity to enforce information consistency, input data are presented in buyer-specific input information, leading to duplicate records, data inconsistencies, and substantial manual intervention. To address this problem, this study adopts the Action Design Research (ADR) approach to design an artefact that embeds an artificial intelligence (AI)-based model in a blockchain, named Engineering-Aware and Standardisation-Integrated Entity Recognition (EASIER). Such an artefact is designed to automatically read and transform input data into the SME's own standardised format and generate data entries in their manufacturing system. We develop and evaluate the proposed artefact using a real SME's BOM dataset, and our results demonstrate that it achieves 99.77% in both accuracy and an F1-score on the final test set. For SMEs, the proposed method can significantly reduce manual intervention effort and improve input data consistency. This study contributes to both theory and practice by demonstrating how AI-enabled standardisation artefacts, supported by blockchain-based governance mechanisms, can be designed with SMEs to deliver practical, resource-efficient digital transformation solutions.
Hyperautonomous AI facilitates medical diagnosis and digital sharing of medical records, whereas data security and privacy protection remain critical issues for smart healthcare. This work proposes an efficient anonymous and traceable certificateless aggregate signature scheme. It removes certificate management overhead and solves the key escrow problem. Pseudonyms and traceability reconcile privacy protection and supervision, and aggregation plus batch verification lower computational and communication costs. Evaluations prove the CASA scheme has stronger security and better efficiency, well suited for privacy-sensitive and resource-limited hyperautomated smart healthcare.
AI-enabled weight and metabolic health management at population scale requires enterprise information systems that transform multimodal data into real-time decision-support workflows. This study proposes the Comprehensive Nutrition and Flexible Caloric Diet (CNFCD) model as a multimodal AI-enabled intelligent hyperautomation framework integrating five human-centered components. Based on real-world operational data from over 110,000 users, an analytical cohort of 18,539 participants generated more than 140,000 longitudinal observations. Results showed significant improvements in body weight, body composition, self-efficacy, and dietary behavior. Key predictors of weight reduction included real-time feedback, slow and mindful eating, and daily green vegetable intake.
Port authorities require interpretable methods to identify sea areas with elevated abnormal ship behaviour risk in intelligent maritime safety systems. This study develops an enterprise system algorithm that integrates fuzzy inference, Fuzzy C-Means, chaos-based logistic mapping, and Marine Geographic Information System visualisation using Automatic Identification System data. The framework constructs grid-level indicators from Course Over Ground, Speed Over Ground, Rate Of Turn, heading, Cross-Track Distance, and drift angle. It produces the Manoeuvring Dynamics Index and Trajectory Deviation Index, then transforms grid scores into chaos-prone risk patterns. The results support hotspot identification, patrol planning, and traffic management.
Virtual streamers have emerged as an alternative to influencer streamers. How to balance virtual streamers' cost-efficiency and influencer streamers' cross-channel information spillover is crucial for manufacturers. We study a dual-channel model including traditional online and live-streaming (LS) channels, where the manufacturer chooses LS modes (virtual or influencer streamer). We show that the virtual streamer is optimal for the manufacturer when the influencer's commission rate is high and the LS channel holds a large market share. Surprisingly, the traditional retailer prefers competing with the influencer streamer, as the virtual streamer intensifies competition by lowering prices, while the influencer streamer differentiates channels.
This research explores the merging of multimodal AI and hyperautomation to develop agentic social robot tutors for educational and enterprise use. It introduces the Empathetic Cognitive Robot (ECRobot), a hyperautomated tutoring system designed for complex problem-solving. Key innovations include an event-driven architecture that manages multimodal inputs with zero data loss and an automated validation pipeline using large language model-based student agents to simulate diverse learner behaviours. Findings show that model capacity significantly impacts learning outcomes, with minimal influence from tutoring style. The approaches provide scalable, robust frameworks for real-time data coordination and stress-testing AI systems across various applications.
Requirement engineering (RE) produces and manages well-defined requirements and is critical in cyber-physical system (CPS) development. Within RE, requirement traceability models fail due to collaboration, trust and privacy issues. The study proposes a decentralised application-based requirement traceability (DART) framework to support CPS development. The DART framework involves an Ethereum blockchain-based decentralised application, an advanced encryption standard (AES) cryptographic algorithm and graph-based representations. The proposed framework can enhance communication and security, positively impacting stakeholder collaboration, trust and decision-making. The privacy feature gives stakeholders confidence in sharing proprietary information, thus enabling continuous and consistent knowledge of the emergent CPS development.
This study integrates Generative AI (GAI) and Design Thinking to improve learners' ability to identify user needs in Enterprise Information Systems. Findings show that satisfaction and involvement are the strongest predictors of learning effectiveness, while subjective norms shape attitudes toward this pedagogy. By extending TRA and TPB, the research contributes to educational technology theory and offers a practical framework for training IT professionals to develop more user-centered, innovative, and effective information systems.
This study proposes a Bi-symmetrical Weighted Distance (BWD) optimization framework for multimodal AI system development under uncertainty. By integrating fuzzy multi-objective linear programming with possibilistic programming, the approach simultaneously minimizes development costs, deployment time, and acceleration costs. The BWD method effectively handles imprecise parameters through distance-based defuzzification, enhancing decision transparency in intelligent hyperautomation contexts. An industrial case study validates the methodology, demonstrating practical capability to navigate trade-offs among time, cost, and resources where traditional methods struggle with uncertain parameter relationships.
Adoption of the industrial metaverse in manufacturing is limited by architectural barriers in Digital Twins (DTs), notably centralised data collection, data-sovereignty conflicts, vendor silos, and weak decentralised governance, which increase privacy risks and hinder collaboration. This paper presents a four-layer framework that combines hierarchical federated learning for privacy-preserving edge intelligence, proof-of-authority blockchain, hybrid physics-informed and data-driven DT simulation, and human-in-the-loop metaverse interaction. Using the NASA C-MAPSS FD001 dataset, the framework achieved an RMSE of 12.49 for RUL prediction, outperformed FedAvg by 23%, remained within 2.8% of centralised training, and cut communication overhead by over 50%.
Enterprise systems in Industry 4.0 increasingly rely on digital twins and metaverse integration for real-time monitoring, control, and decision-making. However, these systems remain vulnerable to evolving cyber threats, and traditional attention-based models used in intrusion detection suffer from high computational complexity (O(n2)). To address this, we propose a sparse attention-based transformer model that used entropy for the calculation of attention. The model combines categorical embeddings with continuous features and applies entropy-guided sparse attention to reduce complexity and enhance feature relevance. Using the CIC-DDoS2019 dataset, the model achieves 95.24% precision and high AUC scores (0.99) while using significantly fewer trainable parameters.
This paper advocates process maps as boundary objects to facilitate communication, coordination and collaboration between enterprise architects and process analysts. Since process maps model business process architecture, they cannot be expressed using BPMN. Although ArchiMate is a potential language, it lacks the expressiveness needed to represent structures such as process chains, groups and families. To address this, the Business Process Architecture Language (BPAL) is presented. A controlled experiment evaluates its usability, and an illustrative scenario demonstrates its boundary-spanning capabilities to support strategically aligned business process architecture.
This research proposes a Metaheuristic Self-Attention Generative Adversarial Network (MetaSeGAN) framework to enhance predictive performance in employee attrition analysis within the context of Human Resource Management (HRM). The MetaSeGAN framework leverages a Conditional Tabular GAN to generate high-quality synthetic tabular data, addressing imbalanced datasets and improving training effectiveness. Furthermore, a self-attention mechanism is integrated into the generator network to capture complex dependencies and relationships within the data, allowing the model to focus on key features that significantly influence employee attrition. Experiments conducted on the Kaggle HR Analytics and New England College of Business datasets demonstrate strong performance, achieving accuracies of 99.10% and 88.80%, respectively, and outperforming baseline methods, including CTGAN, SMOTE, ADASYN, and ForestDiffusionGenerator.
The Metaverse blends physical and digital realities, transforming social, educational, and professional interactions into immersive experiences. However, its rapid expansion introduces risks such as addiction and negative mental health effects. This study explores these concerns, focusing on virtual environment addiction and the need for proactive solutions. It proposes responsible usage, integrated mental health support, and ethical interventions like avatar-based therapy and AI moderation. A case study of 100 BGMI/PUBG players aged 18-25 reveals addiction symptoms, including withdrawal, neglect of responsibilities, and preference for virtual over real-life interactions. The findings highlight the Metaverse's dual impact and emphasize the need for evidence-based design, regulation, and mental health safeguards.
Blockchain technology holds significant potential to transform business operations and generate business value. However, how organisations derive business value from its implementation and use remains unclear. Thus, this study aims to explore how business value is generated through blockchain implementation and use. Drawing on a systematic literature review of 110 studies, the study develops a unified framework outlining key conditions for value realisation, including strategic imperatives, blockchain investments, assets, impacts, and business value. The study contributes to theory by clarifying value creation mechanisms and provides practical guidance for organisations. It also distinguishes between permissionless and permissioned blockchain architectures and explains their roles in value generation.
As Artificial General Intelligence (AGI) is moving from speculation to strategic concern, clear guidance for its responsible adoption remains limited. We propose a human-centered framework for AGI integration grounded in governance, accountability, and regulatory alignment. Drawing on YouTube transcripts from leading industry voices, we develop and test our model through a controlled simulation of human-AI decision-making. Our findings show that AI trained with human rationales improves decision accuracy and strengthens the link between reliance on AI and outcomes. We offer a practical pathway for organizations to adopt AGI while preserving human judgment and ethical responsibility.
Edge-centric digital process twins struggle with slow adaptation and real-time compliance. We propose a self-evolving Edge-AI architecture with five components. The Hierarchical Neuro-Symbolic Verification Graph integrates symbolic rules and neural graphs, reducing latency by 35% while maintaining over 97% compliance. Federated Evolutionary Drift Adaptation improves drift response by 28% and F1 score by 15% using evolutionary operators. The Multi-Resolution Spatio-Temporal Causal Inference Network enables 40% earlier fault detection via causal separation. The Quantum-Inspired Edge Reinforcement Optimiser speeds convergence by 30% and boosts effectiveness by 12%. The Cross-Layer Digital Twin Consistency Ledger ensures tamper-proof state integrity with <0.5% violations at 10k+ transactions/s, enabling secure, adaptive automation.
