AI-Driven Secretarial Workflow Automation for SMEs

Thesis Abstract

The rapid digitization of administrative functions in small and medium-sized enterprises (SMEs) has intensified the demand for efficient, accurate, and scalable secretarial workflows, yet many SMEs struggle with fragmented processes, limited ICT capabilities, and resistance to change that inhibit productivity gains from automation. This study addresses the problem by evaluating how AI-driven workflow automation can transform routine secretarial tasks—document management, scheduling, correspondence, and information retrieval—into integrated, decision-support enabled processes that reduce cycle times and error rates while preserving professional judgment. The aim is to design, implement, and evaluate an AI-enabled secretarial workflow framework tailored to the resource constraints and operational rhythms of SMEs. The specific objectives are (1) to identify key secretarial tasks amenable to AI augmentation and quantify baseline performance metrics (cycle time, error rate, user satisfaction); (2) to develop an integrated AI workflow prototype incorporating natural language processing, robotic process automation, and adaptive scheduling, guided by the theoretical lens of socio-technical systems and the Technology Acceptance Model (TAM); (3) to evaluate the prototype through a mixed-methods study across 12 SMEs over a six-month pilot, measuring changes in efficiency, accuracy, and perceived usefulness; (4) to examine organizational and user-related determinants of adoption, including change readiness, perceived ease of use, and trust in automation; and (5) to formulate a scalable implementation blueprint with governance, data governance, and security considerations for broader SME deployment. The methodology adopts a sequential explanatory mixed-methods design. The initial quantitative phase involves a quasi-experimental pretest–posttest with 12 SMEs, each with a minimum of three secretarial staff, collecting metrics on document turnaround time, meeting scheduling accuracy, and email triage effectiveness (n ? 180 participants). Data collection instruments include system log analytics, task timing dashboards, and standardized user surveys. The subsequent qualitative phase comprises semi-structured interviews (n ? 36) and focus groups with secretaries, IT staff, and managers to contextualize quantitative results and uncover facilitators and barriers. Validity and reliability are ensured through triangulation, instrument piloting, and inter-rater reliability checks for thematic coding. Analytical techniques include descriptive statistics, paired-sample t-tests and repeated-measures ANOVA to assess performance changes, regression analysis to identify predictors of adoption (e.g., change readiness, perceived usefulness, trust), and thematic analysis of interview data aligned with the socio-technical framework and TAM constructs. A conceptual model synthesizing AI automation capabilities with human-in-the-loop governance and SME-specific constraints guides analysis and interpretation. Anticipated findings include statistically significant reductions in document turnaround times (mean reduction ? 25%), improved scheduling accuracy (? 20% reduction in conflicts), and higher perceived usefulness and ease of use post-implementation, moderated by change readiness and trust in automation. The study is expected to reveal nuanced insights into how AI-driven assistants can complement secretaries by handling routine, rule-based tasks while enabling higher-order judgment, personalization, and relationship management. Theoretically, the research contributes to the literature by integrating socio-technical theory with TAM in the context of AI-enabled administrative work, extending the application of dynamic capabilities theory to SME operations, and detailing a practical framework for AI governance in low-resource settings. Practically, it yields an implementable blueprint for SMEs, including process maps, data governance policies, risk mitigation strategies, cost–benefit considerations, and a phased rollout plan aligned with organizational change management. The conclusions are anticipated to advocate for a modular, domain-adaptable AI workflow platform that emphasizes transparency, user control, and continuous learning. Recommendations will address stakeholder engagement, training programs, governance structures, ethical and regulatory compliance, and pathways for scaling from pilot to enterprise-wide deployment across diverse SME sectors.