Design of an AI-enabled Secretariat Workflow System: Implementation and Evaluation

Thesis Abstract

The study investigates the design, implementation, and evaluation of an AI-enabled Secretariat Workflow System to enhance administrative efficiency, decision timeliness, and document governance in modern executive offices, addressing persistent bottlenecks in task routing, meeting management, and records handling. The problem addressed is the fragmentation of manual processes, inconsistent knowledge capture, and limited scalability of traditional secretarial practices in high-demand organizational environments. The aim is to develop a modular AI-enabled workflow platform that automates routine tasks, optimizes scheduling and correspondence routing, and improves compliance through intelligent document management, while ensuring user acceptance and operational viability. Specific objectives include (1) designing an architectural blueprint for an AI-enabled workflow system with modular components for task assignment, calendar integration, document classification, and natural language–driven query handling; (2) implementing a functional prototype integrated with an enterprise ERP/CRM environment and cloud-based storage; (3) evaluating usability, reliability, and performance against predefined benchmarks; (4) assessing impact on process cycle times, error rates, and user satisfaction; and (5) deriving governance and ethical guidelines for AI-assisted secretarial activities. Methodologically, the research adopts a mixed-methods design combining a controlled implementation study with a longitudinal user evaluation. The population comprises executive secretaries, administrative assistants, and records managers across five mid-to-large organizations. A purposive sample of 60 participants will be recruited, stratified by role and prior exposure to digital tooling. The study employs a two-stage data collection approach first, quantitative data through system usage logs, task completion times, error rates, and service-level attainment over a 12-week pilot, complemented by pre- and post-implementation surveys measuring perceived usefulness, ease of use, and trust in automation; second, qualitative data gathered via semi-structured interviews, focus groups, and diary logs to capture contextual use, perceived benefits, and challenges. Instrumentation includes the System Usability Scale (SUS), Technology Acceptance Model (TAM) questionnaires, and a bespoke workflow performance checklist. Validity and reliability will be established through pilot testing, Cronbach’s alpha for internal consistency (target >0.70), and inter-rater reliability for qualitative coding (Cohen’s kappa >0.60). Data analysis will proceed with descriptive statistics and inferential methods, including paired t-tests and repeated-measures ANOVA to examine changes in efficiency and user attitudes, and regression analysis to identify predictors of adoption and performance gains. The qualitative data will be analyzed using thematic analysis guided by Braun and Clarke, with triangulation across sources to ensure robustness. The analytical framework integrates the Diffusion of Innovations theory to interpret adoption patterns, and the Technology–Organization–Environment (TOE) framework to contextualize organizational factors influencing implementation. A conceptual model linking AI-enabled automation components (intelligent routing, automated documentation, and predictive reminders) with process outcomes (cycle time, error rate, and user satisfaction) will be tested. Expected findings indicate statistically significant reductions in average task cycle times and document processing errors, accompanied by improved meeting coordination and records retrieval efficiency. User acceptance is anticipated to rise, with SUS scores surpassing the 68 threshold and TAM measures indicating favorable perceived usefulness and ease of use. Qualitative insights are expected to reveal enhanced role ambiguity reduction, improved knowledge capture, and increased confidence in AI-assisted decisions, balanced by concerns regarding data privacy, change management, and the need for human-in-the-loop controls. The study contributes to knowledge by operationalizing an end-to-end AI-enabled secretariat workflow design, providing empirical evidence on performance gains and user-centered evaluation in real-world settings, and offering a transferable implementation blueprint that aligns with contemporary governance, compliance, and ethical considerations in intelligent administrative systems. Practical implications include guidelines for system architecture, data governance, change management, and metrics for continuous improvement. The main conclusion posits that an AI-enabled Secretariat Workflow System can substantially improve operational efficiency and governance in secretarial functions when designed with modularity, transparent AI components, and strong alignment to organizational processes. Recommendations focus on scaling the prototype across different organizational contexts, refining governance policies for data privacy and accountability, and investing in ongoing user training and change management to sustain adoption and maximize benefits.