Volume 8.
Issue 09.

Modern manufacturing networks operate under tightly coupled production, material, quality, and equipment constraints, making exception management a central challenge for enterprise resource planning systems. Conventional exception handling generally depends on predefined rules, manual investigation, and sequential escalation, which can delay corrective action when manufacturing disturbances propagate across production and supply processes. This research proposes a Generative AI-powered autonomous decision framework for manufacturing exceptions and real-time SAP S/4HANA supply network coordination. The framework conceptually integrates event detection, contextual diagnosis, multi-agent reasoning, constraint assessment, decision generation, SAP transaction orchestration, and continuous feedback. The theoretical foundation is derived from the provided literature concerning machining deformation, residual stress, ring-rolling processes, casing manufacturing, dimensional inspection, and thermal distortion. These studies demonstrate that manufacturing exceptions frequently originate from interacting physical and process variables rather than isolated events. The proposed framework therefore treats exceptions as contextual, cross-functional decision problems rather than simple alerts. A multi-agent generative AI architecture is introduced to coordinate production, procurement, quality, maintenance, inventory, and supply-network reasoning while maintaining SAP S/4HANA as the transactional system of record. The analysis indicates that such an architecture can improve exception prioritization, shorten decision cycles, support explainable recommendations, and coordinate corrective actions across supply-network functions. However, autonomy must remain bounded by authorization policies, process constraints, data quality, and human governance. The study contributes a conceptual architecture linking generative AI reasoning with manufacturing exception management and SAP-integrated operational coordination.

As enterprises generate vast amounts of data in the cloud, via Internet of Things (IoT) devices, artificial intelligence (AI) applications, and digital business operations, the need for scalable governance and analytics has become apparent, highlighting the shortcomings of having data stored and managed in a centralized location. Data silos, governance issues, data quality concerns, and late data insights are common challenges faced by large enterprises that call for the need for more adaptive and distributed data management methods. Data Mesh and Data Fabric architectures are explored and their transformative impact on scalable data governance and enterprise analytics capabilities is looked at in this study. The overview is achieved by using a systematic literature review and comparative analytical approach, which involved the synthesis of evidence from peer-reviewed scientific and technological literature, industry reports, and enterprise implementation studies from the most important scientific and technological databases. The analysis compares both paradigms with regard to architectural principles, governance mechanisms, operational characteristics and organizational implications, and its impact on data accessibility, governance maturity, the scalability of analytics, and the effectiveness of decision making. The results show that Data Mesh can support the scalability of the domain-oriented ownership, data-as-a-product principle, self-service infrastructure, and federated computational governance, which can promote organizational agility and accountability. Data Fabric, on the other hand, extends enterprisewide integration with metadata-driven intelligence, automated data orchestration, knowledge graph technologies and AI-powered governance capabilities. Both architectures are shown to play a crucial role in enterprises today across a range of complex environments, where they help to reduce data fragmentation, enhance management of data quality, speed up time-to-insight and enable advanced analytics projects. Moreover, the study reveals that organizational culture, governance maturity, metadata maturity, and technological interoperability are key factors in the successful implementation. The paper outlines an integrated governance and analytics transformation framework, which integrates the complementary strengths of Data Mesh and Data Fabric. In today's data-driven business landscape, this research work adds to the existing enterprise data management literature by providing a holistic comparative perspective, and by offering practical advice for organizations that are on a journey of analytics transformation aimed at scalable and governance-centric approaches.

Large Language Models (LLMs) are increasingly capable of generating software artifacts from natural-language requirements, programming specifications, and contextual prompts. Although this capability can accelerate enterprise software development, generated code introduces significant concerns regarding correctness, security, behavioral consistency, traceability, and operational dependability. The central challenge is therefore not merely improving code-generation capability but establishing sufficient trust in generated artifacts before their integration into enterprise systems. This research presents a trust-aware conceptual framework for LLM-assisted enterprise application development grounded in verification, validation, traceability, model differencing, process diagnostics, and digital-twin-oriented assurance principles. The methodology synthesizes the supplied literature to define a multi-stage pipeline encompassing requirement interpretation, contextual generation, semantic inspection, verification, behavioral validation, trace analysis, and deployment-oriented trust assessment. The framework treats generated code as an artifact requiring systematic evidence rather than unconditional acceptance. Verification and validation principles provide the foundation for separating syntactic correctness from functional adequacy, while trace-oriented techniques support the identification of behavioral inconsistencies and deviations. Model-driven engineering and digital-twin concepts further contribute mechanisms for maintaining alignment between intended and implemented system behavior. The resulting analysis indicates that trust in LLM-generated enterprise software should be understood as an evidence-based, continuously evaluated property rather than a binary characteristic. The proposed approach provides a research foundation for integrating LLM productivity with enterprise-grade security and dependability requirements.

The rapid adoption of cloud-native applications, distributed services, automated pipelines, containers, workload orchestration, and machine-to-machine communication has resulted in a substantial expansion of non-human identities (NHIs) within enterprise environments. Unlike conventional human identities, NHIs frequently operate continuously, interact programmatically, and require access across multiple cloud platforms, creating governance and security challenges that cannot be adequately addressed through static access-control practices. This research develops a conceptual framework for continuous identity governance and automated threat protection for NHIs in multi-cloud Identity and Access Management (IAM). The proposed approach combines identity discovery, contextual authorization, behavioral monitoring, graph-oriented relationship analysis, risk evaluation, adaptive policy enforcement, and automated remediation. The methodology synthesizes the computational principles evident in the provided literature, particularly distributed graph processing, parallel computation, clustering, data mining, and scalable analytics, and applies them conceptually to NHI governance. The analysis indicates that graph-based identity representations can improve visibility into relationships among workloads, credentials, resources, and permissions, while distributed processing can support continuous analysis at multi-cloud scale. The study further argues that automated governance should be risk-adaptive rather than exclusively rule-based, allowing access privileges and protective controls to respond dynamically to changing identity behavior. The proposed framework contributes a research-oriented foundation for continuous NHI governance while recognizing limitations related to heterogeneous cloud environments, behavioral baselines, false positives, policy conflicts, and computational overhead.

Federated authentication architectures have become a central mechanism for providing scalable access to distributed applications through Single Sign-On (SSO) and Multi-Factor Authentication (MFA). Their security advantage derives from centralized or delegated identity assurance, but federation also creates complex trust relationships among identity providers, service providers, authentication brokers, clients, and authorization components. Consequently, compromise of one critical component can propagate across multiple relying applications. This research presents a structured threat analysis and risk assessment approach for federated authentication architectures, focusing on SSO and MFA workflows. The methodology combines architectural decomposition, STRIDE-oriented threat identification, workflow-level attack analysis, trust-boundary assessment, likelihood-impact scoring, and control-oriented risk prioritization. Particular attention is given to credential theft, token manipulation, replay, session hijacking, identity-provider compromise, MFA bypass, assertion misuse, privilege escalation, repudiation, and availability attacks. The analytical model emphasizes that MFA does not independently eliminate federation risk because authentication assurance can be undermined at protocol, session, recovery, or trust-management layers. The study also incorporates methodological insights from the supplied literature concerning systematic evaluation, robustness, ensemble reasoning, reproducibility, and decomposition of complex systems. The resulting framework provides a practical basis for identifying high-risk authentication paths and strengthening identity federation through layered controls, continuous monitoring, least privilege, resilient session management, and risk-adaptive MFA.

Concrete production in Uzbekistan relies almost entirely on Portland cement, whose cost and carbon intensity are rising, while the country’s gas-processing plants generate a large surplus of elemental sulfur that is sold at very low prices or exported over long distances. This study evaluates whether this by-product can be turned into an economically competitive construction material by replacing Portland cement concrete with modified sulfur concrete in precast elements for aggressive environments. A techno-economic model was developed for 1 m³ of concrete of comparable strength (class B40 Portland cement concrete versus modified sulfur concrete), covering materials, process energy, moulds, CO₂ emissions and a 50-year life-cycle cost including replacements. Input prices reflect 2026 conditions in Uzbekistan, and uncertainty in 18 parameters was propagated by Monte Carlo simulation with 10,000 runs, complemented by one-at-a-time sensitivity and break-even analysis. In the base case the initial cost of sulfur concrete (516.6 thousand UZS/m³) is 1.6% lower than that of Portland cement concrete (525.0 thousand UZS/m³), and its CO₂ emissions are 85% lower (60.6 versus 413.6 kg/m³). Under uncertainty, sulfur concrete is cheaper at the production stage in only 21% of runs, because the cost of the polymeric modifier dominates its binder cost, but it has a lower life-cycle cost in 87.6% of runs, with a mean saving of 133.8 thousand UZS/m³ (10.1%). Break-even occurs at a modifier price of about 20.8 thousand UZS/kg or a sulfur price of about 325 UZS/kg. Channelling 100 thousand tonnes of sulfur per year into concrete would displace about 131 thousand tonnes of cement and avoid about 103 thousand tonnes of CO₂ annually. Sulfur concrete is therefore a promising niche technology whose competitiveness depends mainly on low-cost modifiers and on pricing of carbon emissions.

Generative and Agentic AI (GenAI & Agentic AI) are changing the nature of how AI is used in the organization from a supportive role to a more active partner and teammate. This systematic review investigates the impact of collaboration between humans and AI on product strategy and innovation management and focuses specifically on the shift from AI assistants to increasingly independent AI teammates. This review brings together the research in the field of market and customer intelligence, opportunity identification, strategic prioritization, product road mapping, ideation, new product development, experimentation, new product innovation evaluation, and implementation, supported by AI. It also explores the organizational and behavioral aspects that impact successful human–AI teaming, such as task distribution, the complementarity of human and AI intelligence, calibration of trust, explainability, communication, autonomy of AI, human oversight, accountability, and creativity. The review suggests that when used alongside human expertise, AI can enhance analysis, decision making, speed up the process of innovation, generate ideas and support organizations in learning. The rise in AI autonomy brings a host of potential issues on the table – hallucination, algorithmic bias, overreliance, privacy, IP rights, security, loss of expertise, and agentic behavior are just a few. The review thus proposes a comprehensive Human–AI Teammate Framework that connects human and AI competencies, tasks and organizations, collaboration processes, AI autonomy and decision authority, team results, and governance. The framework emphasizes that product professionals are becoming increasingly important as the AI orchestrators who are coordinating human skills and more autonomous AI systems. Lastly, the review highlights the research gap in relation to Agentic AI, decision authority, calibration of trust, creativity, governance and cross-industry and cross-cultural implementation.

Distributed systems increasingly depend on JSON Web Tokens (JWTs) to support stateless authentication across heterogeneous services, APIs, microservices, and dynamically changing execution environments. Although JWTs provide an efficient mechanism for conveying authenticated claims, token possession alone can create security weaknesses when a token is copied, replayed, or presented from an unintended context. This research proposes an adaptive token binding framework that strengthens JWT authentication by associating token validity with contextual characteristics of an authenticated session or transaction. The framework integrates token binding, contextual verification, adaptive risk evaluation, lifecycle management, and continuous validation into a unified architectural model. Its theoretical foundation is derived from the supplied literature on software evolution, maintainability, scalability, performance measurement, system quality, resilience, and adaptive environments. In particular, the framework extends the conceptual direction of token binding and contextual verification proposed by Ganapathy (2025), while incorporating maintainability and scalability considerations identified across the software engineering literature. The methodology develops a conceptual security architecture and evaluates it analytically against authentication continuity, contextual consistency, scalability, maintainability, and operational resilience. The resulting model indicates that adaptive binding can improve resistance to token replay and contextual misuse while avoiding rigid binding mechanisms that may negatively affect legitimate distributed workflows. The study concludes that adaptive verification should be treated as a continuously managed authentication capability rather than a one-time token-generation feature.

With the rapid development of enterprise data ecosystems, scalable, flexible, high-performance architectures are needed to support advanced analytics and data-driven decision-making. On-premises and single-cloud data warehousing systems can find it difficult to meet the growing need for real-time processing, elastic scalability, cross-platform interoperability, and cost optimization. To meet these challenges, however, cloud-native architectures for data and multi-cloud data warehousing solutions have appeared as game-changers for contemporary enterprise analytics. This study explores how cloud-native architectural best practices, such as microservices, containers, serverless functions, and automated orchestration, can improve enterprise data environment performance and agility. Additionally, the paper examines the advantages of multi-cloud data warehousing approaches, such as the ability to harness the best attributes of various cloud providers, mitigate vendor lock-in threats, and enhance operational resilience. The study draws on a systematic review of the latest academic research, industry reports and enterprise case studies to collate evidence on the effectiveness of cloud-native and multi-cloud strategies for optimising analytical workloads, fast data integration, enhanced governance and AI-powered business intelligence initiatives. The results show significant benefits for organisations that embrace cloud-native data architecture in terms of scalability, deployment time, resource efficiency, and data analysis responsiveness. Multi-cloud data warehousing solutions also help enterprises optimize workload distribution, boost data availability and provide more support for real-time analytics across geographically scattered landscapes. The study finds that cloud-native data architecture can be an essential building block for next-generation enterprise analytics, and multi-cloud data warehousing platforms can deliver the flexibility needed for enterprise digital transformation. The paper provides a broad overview and synthesis of architectural approaches, business value, implementation hurdles and future perspectives for optimizing enterprise data management and analytics, adding to the expanding knowledge base about modern data infrastructure.