The transition from the traditional member-based two-step design approach towards a more holistic structural design methodology based on system behaviour and advanced finite element modelling will be a reality in the following years. The benefits of this new design approach include (i) the capability of evaluating the resistance of structures directly from finite element models without passing through single member checks, (ii) the potential of designing lighter and more economical structures, and (iii) the ability to estimate the probability of failure of the complete structure, which becomes a specific and measurable property. In light of said advantages, direct design methods are currently being implemented in international design specifications for steel structures, although the definition of partial system safety factors that guarantee the target reliability requirements of the codes have not been fully developed yet. The fact that the failure probability of the whole structure can be estimated in bulk, poises this new design method as an excellent alternative for the assessment of existing structures, allowing to plan and optimize retrofitting actions rationally. To achieve such, however, a specific reliability framework that accounts for the variability of the system strength throughout the service life of steel structures is necessary. With the objective of adapting the existing system-based direct design approach and contributing to the development of such reliability framework, this paper investigates and presents a procedure to consider different residual working life scenarios through time-dependent reliability analyses. Underpinned by statistical models for system resistance on a variety of steel structures and load variability models for different reference periods, as well as by suitable target reliability values adjusted to existing structures, this research evaluates the probability of failure of steel structures for different residual working lifetimes using First Order Reliability Methods.