In structural design, it is essential to analyze dynamic external actions and assess the effects and significance of vibrations to prevent issues such as excessive cracking and strong oscillations. These aspects can be controlled through proper structural sizing or by ensuring that the natural frequency remains within acceptable limits. Traditional methodologies for determining natural frequency require specialized technical knowledge, are costly due to the need for advanced equipment and software, and demand significant time for execution. To address these challenges, this study proposes the development of an alternative predictive tool based on Artificial Neural Networks (ANN) to estimate the natural frequencies of concrete slabs in a simplified and cost-effective manner. The network architectures were designed using a database of in situ measurements from ribbed and precast slabs. Supervised training, validation, and testing were conducted using two different architectures. Among the analyzed hyperparameters, early stopping played a crucial role in halting training at the optimal point, minimizing error values according to the selected performance metrics. The trained networks were tested with 30% of the input data, comparing the predictions with actual values. The results indicate that the ANN model effectively predicted natural frequencies for low-frequency slabs. However, for slabs with higher frequencies, the predictions were less accurate, likely due to missing data in the database. This suggests the need for data augmentation techniques or the inclusion of additional experimental measurements to improve model accuracy for high-frequency slabs.