A computationally efficient way of solving large-scale problems in structural dynamics is the multi-time-step (MTS) method, which gives fast and accurate solutions by dividing the spatial domain of the problem into subdomains and enabling the use of different time-steps in different subdomains. With MTS methods one is able to focus the available computational resources on solving the specific spatial regions of interest with a fine resolution while the remainder of the structure is solved with a large time-step to reduce computational cost. To maintain accuracy, however, it is crucial to couple the individual subdomains back together correctly. This is achieved by enforcing continuity constraints on the interface between subdomains using Lagrange multipliers. Conventional MTS methods compute the solution to such problems using a computationally intensive 3-step process: solve the decoupled subdomains individually, compute Lagrange multipliers on the interface, and update the subdomain solutions to satisfy continuity constraints. To circumvent the 3-step solution process of MTS methods, we use a machine learning (ML) approach to predict the Lagrange multipliers on the interface. Solutions generated using MTS methods are used to train an artificial neural network that uses state vectors from the subdomains as input and produces values of the Lagrange multipliers on the interface as output. Since the Lagrange multipliers act as distributed internal forces between subdomains, they may be interpreted as external forcing functions acting on each subdomain. Thus, the response of a particular subdomain due to the effect of the Lagrange multipliers is obtained by simply treating it as an external force on that subdomain and solving it with conventional time-stepping. With this approach, all subdomains are solved and coupled simultaneously in a single pass without the need for a time-consuming 3-step process. Performance of this ML-assisted approach is compared to existing MTS methods in terms of stability, accuracy and computational cost using key numerical examples.