Modern manufacturing techniques produce advanced materials with heterogeneous microstructures, which are challenging to model using phenomenological constitutive laws and calibrated by mechanical tests. To resolve the issue, concurrent micro-macro mechanical simulation is accurate but computationally infeasible. Hence, it is essential to develop AI-enabled material multiscale analysis to fill the gap and accelerate virtual product development for industrial digitalization. In this talk, we will present an AI-enabled material multiscale simulation workflow for industrial-level applications. This AI-enabled workflow is driven by the physics-based deep material network (DMN) which was proposed and developed by LS-DYNA in 2019 for short fiber-reinforced composite (SFRC) materials. The key feature that distinguishes DMN from other neural networks is that it recognizes heterogeneous microstructure patterns in linear elastic training data to make efficient online prediction of nonlinear material behaviours in macroscale. First industrial application of DMN was released in 2022 LS-DYNA for multiscale structural analysis of injection moulded short fiber-reinforced bumper cover. Recent version of LS-DYNA DMN introduces a streamlined simulation workflow for multiscale structural failure analysis accounting for the relationships of process-(micro) structure-property-performance, thereby offering a useful AI-enable multiscale modeling approach for material design and virtual mechanical tests for injection-molded fiber composites. This talk will be concluded by discussing opportunities and challenges beyond DMN short fiber composites analysis.