implementation science

This study examines how Enterprise Architecture (EA) can be localized within Vietnamese government agencies operating under weak institutional foundations to support digital transformation. Addressing EA’s conceptual ambiguity and poor contextual fit, we propose a dual translation mechanism: “theoretical translation”—abstracting indigenous practices into generalizable concepts—and “contextual translation”—deconstructing EA into actionable, organizationally prioritized interventions. Drawing on a 15-year longitudinal case study and integrating mechanism-based analysis with sensemaking theory, we identify critical diffusion pathways for EA in institutionally immature environments. Our findings extend EA theory’s applicability to digital governance in developing countries and yield a reusable conceptual translation framework. This framework offers methodological guidance for digital capacity building across the Global South, bridging theory-practice gaps in public-sector digital transformation.

This study addresses the challenge of endowing traditional business processes with intelligent reasoning and adaptive capabilities while preserving the determinism of existing workflow engines. To this end, the authors propose a “workflow suite” mechanism that enables dynamic intervention by embedding a layer of policy-constrained agents at critical control points. They introduce a novel Task-Decision-Flow (TDF) model that defines three types of collaborative agents and integrates the FRAME policy framework to govern large language model (LLM) invocations, thereby harmonizing structural compliance with normative autonomy. Leveraging a hook-based integration architecture, the approach is implemented and validated within the CUGA FLO system using a loan approval case study, demonstrating a balanced synthesis of process determinism and intelligent flexibility.

This work addresses the limitations of existing evaluation methods, which focus narrowly on task completion and fail to ensure trustworthy deployment of embodied agents in multi-step, externally impactful scenarios, while also lacking coordination among evaluation, governance, orchestration, and runtime assurance. To bridge this gap, the paper proposes an integrated four-layer framework that establishes, for the first time, a closed-loop mechanism linking governance obligations to verifiable execution. Guided by the ODTA principles—Observability, Decidability, Timeliness, and Attestability—the framework introduces runtime localization testing and minimal action evidence bundles. Through a human-in-the-loop evidence synthesis approach, it formally connects policy requirements to concrete agent behaviors, exposing critical gaps such as the inability of static permissions and prompts to govern path-dependent actions. Validation via an enterprise procurement agent demonstrates the framework’s capacity to unify safety, robustness, and trajectory-level evaluation.

This study addresses the inadequacy of the current U.S. Department of Defense software acquisition pathways in effectively managing the unique challenges posed by artificial intelligence systems—particularly their data dynamism, model evolution, and governance requirements. Through scenario-based policy analysis, the authors embed a hypothetical AI-enabled project into critical junctures of the existing acquisition process to systematically evaluate how policies translate into practice. The analysis reveals that core guidance documents lack operational specificity, while AI-related controls are fragmented across supplementary materials, leading programs to rely on inconsistent local interpretations. To bridge this gap, the paper proposes a dedicated AI acquisition sub-pathway alongside targeted documentation enhancements, substantially aligning policy with practice in areas such as data provenance, lifecycle management, and human oversight.

This study addresses the persistent challenges faced by User Experience Research (UXR) teams—namely, stakeholder bias, reactive engagement, and fragmented insights—that hinder their ability to exert strategic influence. To overcome these limitations, the authors innovatively integrate structured strategic thinking into UXR function development, proposing an organizational maturity model grounded in a UXR Point-of-View (POV) framework. Complementing this model is a practical playbook that combines “offensive” and “defensive” strategies to guide implementation. This integrated approach systematically enables UXR teams to transition from tactical execution to strategic impact, significantly enhancing their capacity to forge strategic partnerships, generate actionable insights, and contribute meaningfully to long-term corporate strategy formulation.

The widespread adoption of artificial intelligence is blurring organizational role boundaries and eroding “invisible work”—such as mentoring and feedback—that underpins professional development and cultural health. Through semi-structured interviews with 24 product professionals in technology firms and subsequent thematic analysis, this study systematically uncovers AI’s dual impact: while enhancing peer-level collaboration, it simultaneously weakens traditional mechanisms of career support. To address these tensions, the research introduces a strategic framework that renders invisible work visible and offers actionable interventions for organizations, leaders, and individuals. These measures aim to preserve cultural sustainability without compromising operational efficiency in AI-integrated workplaces.

This study addresses the frequent failure of enterprise technology modernization initiatives due to the absence of structured governance mechanisms. Building on 24 years of practical experience, the authors propose the EMRGF framework—an end-to-end integrated model encompassing governance of cloud and legacy systems, data platform reliability, AI-driven automation governance, and root-cause analysis for mission-critical operations. EMRGF pioneers a standardized approach to cross-domain governance spanning migration, data platforms, and AI, aligning with NIST CSF 2.0, NIST AI RMF, and U.S. Executive Orders 14028 and 14110. The framework integrates four interlocking modules, five implementation tool categories, and a trainer development mechanism. Empirical validation demonstrates that its large-scale adoption reduces development effort by 30%, shortens testing cycles by 35%, enables zero-downtime high-load data migration, and achieves 99.9% reliability in critical analytics pipelines.

This work addresses the absence of benchmarks evaluating language model agents’ ability to consistently adhere to complex, constraint-laden instructions—such as corporate policy manuals—over long contexts and multi-turn tool interactions. The authors introduce the first benchmark for this challenge, comprising 65 tasks across five professional domains, which requires agents to operate within a simulated office environment (e.g., email, calendar, chat) guided by dynamic policy manuals ranging from 20 to 124 pages. Leveraging expert-authored, non-redundant manuals and 824 deterministic scoring rules, the benchmark enables fully automated, stringent evaluation where all criteria must be satisfied. Experiments reveal that even the best-performing configuration among 30 state-of-the-art models passes only 36.2% of tasks, with most scoring below 25%, exposing systemic deficiencies in policy compliance and behavioral consistency.

This study addresses the persistent challenge in software engineering research of empirically validating theories due to the absence of systematic, reproducible operationalization methods. To bridge this gap, the authors propose an integrated methodological framework that combines Sjøberg’s operationalization approach with Dubin’s theory-building methodology, offering the first evidence-driven and replicable guide for operationalizing theoretical constructs in software engineering. The approach systematically translates abstract theories into measurable forms by rigorously defining variables, selecting appropriate indicators, and deriving non-causal assumptions. The utility of the framework is demonstrated through its application to a theory on DevOps team classification. The resulting methodology provides researchers with a robust foundation for conducting verifiable theoretical studies while simultaneously offering practitioners actionable, theory-informed insights.