Tekoälyn soveltaminen hintojen ja toimitusaikojen ennustamiseen hankintatoimessa

Global competition and growing market uncertainty have increased the challenges of supply chain management, highlighting the importance of data-driven forecasting in procurement. Although companies have more data available than ever, utilizing it to forecast future trends is often challenging, and processes still rely largely on traditional statistical methods and experience-based judgement. The aim of this thesis was to explore the use of artificial intelligence in forecasting electronic component prices and lead times, and to develop a forecasting tool to support procurement professionals in price negotiations. The study also examined the key requirements and challenges of AI-based forecasting. The study was conducted using the design science research methodology. The empirical part consisted of expert interviews with the case company, which were used to define the requirements for the solution. Based on this, a Python-based hybrid model was designed and implemented, combining machine learning and deep learning methods. The developed solution utilizes both component-specific market data and external macroeconomic indicators. The result of the work was a functional Proof-of-Concept solution that produces 12-month forecasts and visualizes the factors affecting them to improve the interpretability and transparency of the forecasts. The results showed that the AI-based approach offers a structured and repeatable alternative to experience-based forecasting. Although the limited historical data available restricted the full validation of the model's long-term accuracy, the solution proved to be technically valid and scalable. The conclusion is that the successful use of artificial intelligence requires sufficient data quality and quantity, as well as explainability to gain user trust. The work provides a foundation for more data-driven forecasting in procurement as longer historical time series become available.