In this contribution, the authors will present two cases where AI-based approaches are useful in constitutive modelling and model parameter determination for the simulation of composite cable structures. The determination of model parameters for simulation tasks in the automotive industry including flexible objects such as cable bundles is tedious even for the simplest linear-elastic constitutive model since variations in the composition of cable bundles in applications are high. The authors will present an estimation approach using Gaussian Process regression for data-based estimation of effective bundle parameters which yields a predictive distribution of the desired parameters, using easily accessible bundle characteristics as inputs. Electric cables are complex objects due to their multi-material composition and geometric properties. These characteristics lead to various internal interaction effects, resulting in observed inelastic and hysteretic deformation behaviour. Hysteresis operators of the Preisach type, which can be expressed as a weighted superposition of simpler operators, are used to describe input-output relations in hysteretic behaviour. Here, this approach is used to model the moment vs. curvature relationship for cable bending. Such operators can be incorporated into a Cosserat rod model as inelastic constitutive laws for quasistatic simulations.