AI-powered Real-Time Secretariat Workflow Optimizer for SMEs

Thesis Abstract

The rapid digitization of secretarial functions in small and medium-sized enterprises (SMEs) has intensified the demand for integrated, real-time workflow coordination to enhance productivity, reduce administrative bottlenecks, and improve decision-making timeliness. Despite advances in enterprise resource planning and AI-enabled assistant tools, SMEs still rely on ad hoc processes and siloed information flows that impede responsiveness. This study addresses the problem by developing and evaluating an AI-powered Real-Time Secretariat Workflow Optimizer (RT-SWO) designed to orchestrate scheduling, document handling, meeting management, and record-keeping across multiple secretarial tasks. The aim is to investigate how real-time AI-driven orchestration influences operational efficiency, task accuracy, and stakeholder satisfaction within SME secretariats. Specific objectives are (1) to design an architecture for RT-SWO integrating natural language processing, machine learning-based task routing, and event-driven microservices; (2) to evaluate the impact of RT-SWO on cycle time reduction and error rates in secretarial workflows through a quasi-experimental study; (3) to examine user perceptions of system usefulness, ease of use, and trust in AI automation; (4) to identify implementation challenges, data governance considerations, and security implications; and (5) to develop a scalable deployment guideline tailored to SME contexts. The methodological approach adopts a mixed-methods design anchored in the Technology Acceptance Model (TAM) and the Diffusion of Innovations (DOI) framework. The population comprises 60 SME secretariat units across three sectors (professional services, manufacturing, and retail) within a metropolitan region. From this population, a purposive sample of 40 secretarial staff and 12 managers will participate in a 12-week intervention, with 20 staff and 6 managers serving as a control group operating with conventional workflows. Data collection instruments include (i) time-motion logs and system-generated metrics (cycle time, task completion rate, rework rate, and document turnaround time), (ii) an AI-augmented workflow audit checklist, (iii) a standardized survey measuring perceived usefulness, perceived ease of use, trust in automation, and job satisfaction, and (iv) semi-structured interviews to capture qualitative insights on user experience and governance concerns. Validity and reliability will be ensured through instrument triangulation, pilot testing with 5 participants, and Cronbach’s alpha checks (>0.7) for survey scales. Quantitative analysis will employ difference-in-differences (DiD) to assess changes in cycle times and accuracy between intervention and control groups, complemented by multiple linear regression to control for confounders such as firm size and prior IT maturity. Time-series analysis will examine long-run trends in workflow metrics. The qualitative component will use thematic analysis of interview transcripts to identify perceived value, usability barriers, and data governance issues, with codes converged through inter-coder reliability checks (Cohen’s kappa > 0.75). System logs will be analyzed to identify patterns in task routing efficiency, with feature importance assessed via SHAP values to elucidate the AI decision-making process. A risk assessment will address data privacy, security, and compliance with pertinent regulations, and a cost-benefit model will estimate ROI under varied adoption scenarios. Expected findings include statistically significant reductions in average document turnaround time and task cycle duration, improved accuracy in meeting scheduling and record maintenance, and higher user satisfaction among participants in the intervention group. The study also anticipates nuanced insights into how trust in AI, perceived ease of use, and organizational readiness mediate implementation outcomes. Theoretically, the research will extend TAM and DOI by integrating real-time operational metrics with human-AI collaboration dynamics in procedural secretariat contexts, contributing a refined model of AI-assisted workflow orchestration for SME environments. The anticipated contribution to knowledge encompasses (i) a validated architectural blueprint for real-time AI-driven secretariat workflow orchestration adaptable to diverse SME settings, (ii) empirical evidence on efficiency, accuracy, and user acceptance impacts of RT-SWO, (iii) governance and security guidelines for data flows across confidential documents and meeting records, and (iv) a practical deployment framework detailing integration steps, change management, and ROI articulation. Policy implications include recommendations for SME digital literacy programs and vendor selection criteria emphasizing interoperability and data sovereignty. The study concludes that RT-SWO can substantially enhance secretariat performance in resource-constrained SMEs when coupled with appropriate governance structures and user-centered design, and it recommends phased scaling, continuous monitoring, and ongoing training to sustain benefits.