Impact of AI-assisted document workflows on office productivity: an empirical field study

Thesis Abstract

The rapid digitization of office workflows has heightened the need to understand how AI-assisted document processes influence productivity in contemporary organizations, where manual bottlenecks and repetitive tasks persist despite automation investments. This study addresses the problem of inconsistent productivity gains from AI-enabled document workflows by examining how adoption, usage patterns, and perceived value translate into measurable performance outcomes in real-world office settings. The aim is to quantify the relationship between AI-enabled document workflows and office productivity, while elucidating the mediating roles of user proficiency, task complexity, and organizational support. Specific objectives are (1) to assess the extent of AI-assisted document workflow adoption across knowledge-intensive offices; (2) to determine the impact of AI usage intensity on objective productivity metrics (document turnaround time, error rates, and throughput) and subjective productivity perceptions; (3) to identify moderating effects of staff digital literacy and training; (4) to explore employee experiences and contextual factors through qualitative insights; and (5) to formulate a validated framework linking AI-enabled workflows to productivity outcomes. A mixed-methods research design is employed, integrating a cross-sectional quantitative survey with a longitudinal field component and qualitative interviews. The study population comprises 12 large and mid-sized organizations across finance, legal, and professional services sectors in a metropolitan region. A stratified random sample targets 420 office workers who regularly engage in document-intensive tasks, with a planned follow-up of 6 months to observe temporal changes. Data collection instruments include an AI-Workflows Usage Survey (validated parasitic construct scales for adoption, usefulness, and ease of use) and objective performance records (document processing time, error frequency, and compliance rates) extracted from organizational workflow systems, complemented by minutes-to-delivery metrics. In-depth semi-structured interviews with 30 participants provide contextual understanding of workflow integration, governance, and perceived value. Reliability and validity procedures entail Cronbach’s alpha checks (? ? .70), factor analyses confirming construct validity, pilot testing with 40 respondents, and triangulation of survey data with system logs. Quantitative data will be analyzed using multiple regression to estimate the effect of AI usage intensity on objective productivity metrics, controlling for job role, tenure, and sector. Hierarchical linear modeling (HLM) will test cross-level interactions between individual-level usage and organizational support. Mediation analyses will explore whether perceived usefulness and ease of use mediate the AI-productivity link. Moderation by digital literacy and training will be examined via interaction terms. Time-series analysis on the longitudinal component will detect trends in productivity following AI workflow adoption. Qualitative data will undergo thematic analysis, with coding guided by the Technology Acceptance Model and the Resource-Based View, to uncover mechanisms, barriers, and enabling practices. A convergent mixed-methods approach will integrate quantitative results with qualitative themes to yield a comprehensive understanding. Key anticipated findings include (i) significant improvements in document turnaround time and overall throughput associated with higher AI usage intensity, moderated by user training; (ii) reductions in error rates contingent on system integration quality and governance structures; (iii) positive correlations between perceived usefulness/ease of use and productivity gains; and (iv) context-specific factors—such as organizational culture and leadership support—that influence adoption efficacy. The study is expected to contribute to knowledge by bridging theory and practice in AI-enabled workflows, advancing a validated conceptual model that links technology, human factors, and performance outcomes. It will offer empirical evidence on the conditions under which AI-assisted document workflows yield meaningful productivity gains, informing managers about optimal implementation strategies, training requirements, and governance mechanisms. The main conclusion will emphasize that AI-assisted document workflows can enhance office productivity when integrated with comprehensive change management, continuous skills development, and robust data governance. Practical recommendations include targeted training programs to raise digital literacy, standardized onboarding for AI tools, and ongoing monitoring of process performance using system-generated analytics to sustain gains.