Reliability-based designs have seen extensive use in structural engineering but face slower adoption in geotechnical practices due to the complexities in understanding uncertainties in the ground. This paper aims to bridge geotechnical and structural designs in sophisticated engineering systems, focusing on shallow square foundations that transfer structural loads to the underlying ground. The study introduces a semi-analytical method to assess the bearing capacity failure probability of shallow square foundations on 3D undrained cohesive soils using local average theory. The failure probability is efficiently calculated through fast evaluations of variances and covariance of locally averaged soil properties without relying on time-consuming computations via 3D non-linear random finite element analysis (RFEA). Confirmed through Monte Carlo simulation validation, the approach accurately determines the necessary resistance factors for designs against bearing capacity failure. Parameters specific to each site, such as sample location, variation coefficient, and fluctuation scale of spatially variable shear strength, influence the resistance factors. The approach suggests a practical method that aligns resistance factors with reliability theories, aiding engineers in transitioning to reliability-based designs without the need for extensive computations.