The evolution of artificial intelligence (AI), machine learning (ML), and neural networks (NN) is transforming the landscape of scientific and engineering modeling. It also prompts a debate on the role of first-principles modeling (FPM) in chemical engineering. While data-driven methods excel at interpolation and very rapid development, they often lack physical fidelity, interpretability, and reliable extrapolation capabilities. This article provides a personal academic and industrial perspective on the synergistic integration of FPM and AI-based methods, highlighting their complementary roles in process systems engineering. We argue that FPM (based on fundamental conservation laws and mechanistic understanding of phenomena), remains indispensable for ensuring robustness, safety, physical consistency, and adaptability of models in PSE. Moreover, we analyze the synergistic potential of hybrid approaches by deconstructing the model-building workflow. The latter is the primary lens to identify key decision points where integration delivers maximum value, moving beyond a simple paradigm comparison. Using this structured analysis of the model-building workflow, we identify several major opportunities for this integration, particularly where first-principles knowledge is incomplete. The discussion extends to practical strategies for model validation, scalability, and industrial applications, supported by case studies, as well as the potential of LLMs in assisting the future developments of FPM. Finally, we conclude that a physics-informed foundation for modeling is not obsolete but is instead critical for guiding the safe and reliable application of AI in chemical engineering.
Kiss et al. (2026) studied this question.