Abstract
Efficient workflow management is crucial for the agility and productivity of modern enterprises, particularly in dynamic and complex operational environments. However, traditional workflow systems often suffer from rigidity, lack of adaptability, and insufficient utilization of artificial intelligence (AI) to address evolving business needs. This research addresses these limitations by proposing a deep learning-powered framework for workflow optimization tailored for next-generation enterprise solutions, leveraging the capabilities of workday extend and seamless AI integrations. The framework employs a farmland fertility optimization (FFO) algorithm to fine-tune hyper-parameters and improve the convergence of deep transfer learning models. The convolutional neural network (CNN)-based architectures are utilized to capture complex patterns and dependencies in workflow execution data, enabling intelligent decision-making for dynamic process adjustment. The proposed methodology is validated on the publicly available enterprise workflow dataset (CRM), which encompasses diverse real-world workflow logs from multiple industries. Comparative analysis shows that the integration of FFO with CNN models achieves key performance improvements, including an AUC increase to 0.978, an F-measure of 0.972, Recall of 0.980, and Precision of 0.965-representing up to 53% improvement over traditional methods and notable gains over baseline CNN models.