The Use of Artificial Intelligence in Financial Forecasting and Strategic Decision Quality: The Mediating Role of Perceived Forecast Quality in Moroccan Firms

Abstract

As finance functions digitalize, firms increasingly integrate artificial intelligence (AI) tools into forecasting, yet research on Morocco has mainly examined the predictive performance of models and says little about how forecasts are used in decisions. This study addresses this gap by examining whether AI use in financial forecasting and data quality are associated with perceived forecast quality, and whether the latter is associated with strategic decision quality. It contributes by specifying perceived forecast quality as a statistical transmission mechanism between AI use and decision quality. The study adopts an embedded mixed-methods design with quantitative dominance: ten exploratory interviews provide contextual framing, followed by a non-probability survey of 250 finance and management-control professionals in Moroccan firms, analyzed with PLS-SEM. Results show positive associations of moderate to small magnitude: data quality (β = 0.259) and, to a greater extent, AI use (β = 0.364) are associated with perceived forecast quality (R² = 0.250), which is in turn associated with decision quality (β = 0.334; R² = 0.207). The indirect effect of AI use on decision quality is significant (β = 0.121; 95% CI [0.070, 0.183]) alongside a significant direct effect (β = 0.197), indicating complementary partial statistical mediation. Interviews and an open-ended question point to data governance, human oversight, and user support as enabling conditions, which the model does not test. The study thus shifts attention from algorithmic performance alone to the organizational appropriation of AI-assisted forecasting. The cross-sectional design, the perceptual measurement of forecast quality, and the incomplete validation of the items call for caution in interpreting the findings.

JEL Classification: C45, C53, D81, M15, M41, O33.

Paper type: Empirical Research.