AI-Driven Financial Forecasting in SMEs for Strategic Decision-Making

Thesis Abstract

This study investigates the potential of AI-driven financial forecasting to enhance strategic decision-making in small and medium-sized enterprises (SMEs) facing volatility in revenue, liquidity, and investment priorities. The problem addressed is the limited access to accurate, timely, and granular forecast information in SMEs, which constrains proactive planning and resource allocation. The aim is to develop and validate an AI-enabled forecasting framework that improves forecast accuracy, uncertainty quantification, and scenario analysis to support strategic choices such as capital budgeting, working capital management, and revenue diversification. Specific objectives are (1) to design an AI forecasting model integrating machine learning techniques with traditional time-series methods; (2) to evaluate the model’s predictive performance against baseline econometric benchmarks across diverse SME sectors; (3) to assess the impact of forecast outputs on strategic decision quality through decision support experiments; (4) to examine managers’ adoption intents and perceived usefulness of AI forecasts; and (5) to identify governance and ethical considerations in deploying AI for financial planning. The methodology adopts a mixed-methods, explanatory design. A multi-site cross-sectional survey will target 320 SME managers across manufacturing, retail, and service sectors, complemented by 40 in-depth interviews with financial managers and owners to capture contextual factors. Data collection uses a structured questionnaire to obtain financial performance indicators, forecast accuracy, and decision outcomes, alongside semi-structured interview guides exploring adoption determinants and governance considerations. The AI forecasting framework combines an ensemble of gradient boosting machines and recurrent neural networks with traditional ARIMA models, calibrated on quarterly financial data from three-year windows. The population comprises SMEs with annual revenues between $2 million and $50 million, and a stratified sampling approach ensures sectoral representation. Model validation employs rolling-origin evaluation, comparing metrics such as RMSE, MAE, and MAPE, and probability calibration for uncertainty intervals. To assess decision quality, a simulated decision environment tests capital budgeting and liquidity planning under forecast-driven scenarios, analyzed via regression analysis and ANOVA to detect improvements relative to baseline forecasting. The qualitative strand utilizes thematic analysis of interview transcripts, anchored by the Technology Acceptance Model (TAM) and the Diffusion of Innovations theory to interpret adoption dynamics. Expected findings indicate that the AI-enabled forecasting framework yields statistically significant improvements in forecast accuracy (reductions in RMSE and MAPE by 12–18% on average) and tighter prediction intervals, which in turn enhances the precision of liquidity planning and capital allocation decisions. The study anticipates a positive effect of forecast quality on perceived decision usefulness and managerial confidence, moderated by organizational absorptive capacity. Sectoral differences may emerge, with higher gains in retail and manufacturing SMEs due to greater revenue volatility and inventory dynamics. Qualitative insights are expected to reveal critical success factors, including data governance maturity, model interpretability, and alignment with strategic planning cycles, as well as ethical guardrails for data privacy and model bias. The study contributes to knowledge by extending AI-based forecasting literature to the SME context, linking forecast accuracy to strategic decision outcomes, and integrating behavioral adoption theories with technical performance. It offers a practical, scalable forecasting framework tailored for SMEs, with a clearly delineated implementation roadmap, data governance guidelines, and an evaluation protocol that can be replicated across regions and industries. The theoretical contribution lies in operationalizing a hybrid forecasting model within a decision-support ecosystem grounded in TAM and Diffusion of Innovations, providing empirical evidence on how technology acceptance interacts with forecast-driven strategic processes. Policy and managerial implications address how SME ecosystems can leverage AI to improve resilience, optimize working capital, and sustain growth in environments characterized by uncertainty. The study concludes that AI-driven financial forecasting materially enhances strategic decision-making in SMEs when coupled with robust data governance, transparent model explanations, and integrative planning timelines, and it recommends iterative deployment, stakeholder involvement, and continuous performance monitoring to maximize value.