AUK Digital Repository
American University Kyiv electronic data repository, also called an e-archive or centralized data repository
Communities in DSpace
Select a community to browse its collections.
Recent Submissions
AI-AUGMENTED PRE-SALES: DESIGNING AND EVALUATING AN AGENTIC RFP WORKFLOW IN HEALTHCARE AND LIFE SCIENCES
(Manuscript, 2026-05) Budnichenko, Svitlana
This study investigates the design and evaluation of an AI-augmented Request for Proposal (RFP) response workflow within the Healthcare and Life Sciences (HC&LS) pre-sales practice of a global IT services firm. Using a Design Science Research methodology combined with a quasi-experimental before-and-after evaluation, the study develops a Minimum Viable Prototype (MVP) comprising three coordinated AI layers — Research Agent's Cluster, Prototype Brief Writer, and Proposal Narrative Writer agent — embedded within a multi-agent full architecture and
orchestrated via the n8n workflow automation platform.
The workflow incorporates Retrieval-Augmented Generation (RAG) using a Supabase pgvector knowledge base constructed from 192 classified HC&LS proposals from the organization's knowledge base, of which 32 are fully enriched with ten structured fields and indexed via cosine similarity search. The MVP reduced the active effort required to produce a first reviewable draft by approximately 80–85%, compressing the elapsed cycle from a typical 5–7 day process to same-day draft availability, while achieving a proposal quality score of 25/30 (83%) on a six-dimension evaluation rubric.
Findings show that AI creates the greatest pre-sales value when embedded in a redesigned end-to-end workflow as an augmentation layer, not when applied as isolated task automation. Six design principles are derived: augmentation over automation, RAG-based knowledge grounding,
early-stage AI prototyping, modular agent architecture, human-in-the-loop validation, and incremental adoption. The study contributes both a functional prototype and a generalizable framework for AI-enabled workflow redesign in pre-sales contexts.
AI INTEGRATION FOR TEST CASES GENERATION AND MAINTENANCE: OPTIMIZING TEST TEAM WORKFLOW
(Manuscript, 2026-05) Stepaniuk, Taras
Manual quality assurance workflows in fast-paced software delivery increasingly struggle to keep pace with rapid code evolution, with QA teams spending disproportionate effort on interpreting requirements and maintaining test documentation. This research investigates whether an Artificial
Intelligence-driven middleware can optimize this workflow by automating the synthesis of requirements, design, source code, and existing test documentation into actionable testing artifacts.
A middleware solution was designed and implemented on the Fastlane framework, integrating data from Asana, Figma, GitLab, and TestRail, and leveraging the OpenAI GPT-4.1 model through a structured Chain-of-Thought prompt. The evaluation combined quantitative KPI tracking across 24
production tasks with qualitative feedback from three QA engineers.
The results demonstrate a 68% reduction in the test documentation effort ratio, a decrease in the median number of dev/test iterations from three to one, and a doubling of the single-iteration resolution rate. Qualitative analysis confirmed accelerated feature comprehension and reduced cognitive load. The study validates a hybrid human–AI model of quality assurance and defines a roadmap toward autonomous test maintenance.
EVALUATION OF MULTI-BRAND STRATEGY EFFECTIVENESS IN NICHE E COMMERCE UNDER SPONSORED SEARCH AUCTIONS: EVIDENCE FROM CONTRIBUTION MARGIN ANALYSIS
(Manuscript, 2026-05) Shukhrov, Dmytro
The growth of digital platforms allows e-commerce firms to scale via multi-brand models. This study empirically evaluates the financial effectiveness of Company X’s intuitive launch of three custom-printed wallpaper brands with overlapping catalogs but distinct price positioning. It analyzes whether these brands cannibalize each other's demand across three geographic markets. The research addresses four key questions regarding financial impact, source of growth, cannibalization risks, and market-size moderation, testing two main hypotheses (H1 and H2).
Key Findings and Contributions: H1 confirmed: Co-owned brands do not experience demand cannibalization due to algorithmic audience segmentation (distinct seed audiences from different launch timing) and natural price sensitivity differences. H2 confirmed: The multi-brand model generates an additive, not substitutive, increment in total contribution margin. New Concept: Strategy effectiveness depends on the incumbent brand's maturity stage. Launching a second brand prematurely in early-stage markets with increasing marginal returns reduces short-term efficiency. Managerial Recommendations: Company X should use maturity indicators (advertising slope, YoY margin dynamics) to trigger new launches and invest in content differentiation.
THE IMPACT OF SEARCH QUALITY ON E-COMMERCE SALES
(Manuscript, 2026-05) Myropolskyi, Mark
This capstone project examines the impact of internal search performance on customer behavior and e-commerce outcomes using real search data from EVA.UA. Internal search is treated not only as a navigation tool, but also as a mechanism that shapes product discovery, customer experience, and the allocation of economically valuable traffic. The study aims to understand whether query popularity, query characteristics, platform differences, or traffic concentration across demand segments drive search-related business performance.
The empirical analysis is based on a large real-world dataset that includes more than 42 million search events and more than 6 million unique queries, as extracted from the original raw data. After data cleaning, the study applies query-level and event-level analysis to evaluate search demand distribution, conversion patterns, add-to-cart behavior, platform differences, weighted and unweighted performance metrics, frequency deciles, and revenue concentration proxies. Numeric SKU-like queries and low-quality noise were filtered out to reflect genuine customer intent better.
The results show that search traffic is highly concentrated in a small subset of high frequency queries. The top frequency decile accounts for a disproportionate share of total search traffic and revenue proxy. However, conversion differences across frequency segments are relatively small, which suggests that revenue concentration is driven more by traffic allocation than by substantially better query performance. The analysis also shows that query length has a non-linear relationship with conversion: medium-length queries tend to perform best. In contrast, very long queries yield unstable results due to a low sample size. Platform analysis reveals meaningful behavioral differences: web demonstrates the highest purchase conversion, iOS shows the highest add-to-cart rates, and Android underperforms on both conversion and downstream funnel efficiency.
The findings suggest that the major business opportunity is not limited to further optimizing already dominant queries, but also to improving the exposure and handling of underutilized demand, especially on mobile platforms. These results contribute to understanding how internal search influences e-commerce performance and offer practical implications for search optimization, product discovery strategy, and customer
experience management.
ISO 13485:2016 IMPLEMENTATION AND INTERNAL AUDIT READINESS IN A UKRAINIAN MEDICAL DEVICE MANUFACTURER
(Manuscript, 2026-05) Zabroda, Daryna
This capstone examines the implementation of ISO 13485:2016 quality management systems and internal audit readiness within a Ukrainian medical device manufacturer. The research addresses the problem of the gap between formal certification and actual operational effectiveness of quality
management systems in emerging regulatory environments.
This research is guided by the following research questions: how the level of ISO 13485 implementation affects regulatory readiness and compliance performance; how internal audit practices influence the recurrence of nonconformities; and how management involvement impacts CAPA effectiveness.
The aim of the research is to evaluate the relationship between implementation depth, audit performance, and the effectiveness of corrective action.
A mixed-methods single-case study design was applied. Quantitative data were collected through a clause-based gap analysis and internal audit metrics, while qualitative data were obtained through structured document analysis of QMS documentation. The sample consists of one Ukrainian medical device manufacturing organization in the pre-certification stage.
The results indicate a moderate level of ISO 13485 implementation (70.5%), with stronger performance in documentation and weaker integration in monitoring and improvement processes. Internal audits effectively identify nonconformities but have a limited impact on preventing recurrence (32% repeat rate). CAPA processes demonstrate moderate effectiveness, with delays and incomplete closure. The findings suggest that implementation depth, rather than certification status, determines regulatory readiness and quality performance. Management involvement is critical for improving CAPA effectiveness and reducing recurring issues.
The research contributes to theory by emphasizing implementation maturity and provides practical recommendations for improving QMS effectiveness in Ukrainian medical device organizations.