AI and Macroeconomic Modelling: Predicting Currency Exchange Rates and Evaluating Forecast Accuracy as a Business Tool

Artificial intelligence is increasingly integral to macroeconomic analysis as digital transformation reshapes decision-making in management and economics. Its ability to process vast amounts of structured and unstructured data positions it as a potential complement or alternative to expert-driven forecasting. While recent literature documents the predictive capabilities of machine learning techniques, systematic comparisons between advanced language models and institutional macroeconomic forecasts remain scarce, particularly concerning differences in reasoning frameworks and interpretative approaches. This paper examines the application of artificial intelligence in exchange rate forecasting, comparing its performance with that of professional macroeconomic experts. Specifically, three advanced language models are evaluated in forecasting the quarterly Czech koruna to euro exchange rate from 2017 to 2025. A unified ex ante quarterly forecasting design ensures comparability. Predictive accuracy is assessed using root mean squared error and mean absolute percentage error, while qualitative analysis explores the structure and logic of explanatory outputs. The study addresses two research questions: whether artificial intelligence can achieve forecasting accuracy comparable to institutional benchmarks and whether different models exhibit distinct epistemological patterns. Results indicate that all models achieve accuracy comparable to institutional forecasts, with errors below one Czech koruna and percentage deviations around three percent. Concurrently, systematic differences emerge in deductive structure, methodological explicitness, and interpretative restraint. This paper contributes by integrating quantitative accuracy evaluation with qualitative epistemological analysis, advancing methodological discourse on artificial intelligence’s role in macroeconomic modelling and policy-oriented forecasting.