AI-based selection was applied to improve the detection of breast cancer in population-based screening, particularly focusing on interval cancers often missed by mammography. AISmartDensity, a modular AI tool, was used to analyze mammograms and identify individuals at high risk based on three factors: underlying cancer risk, masking potential, and signs of malignancy. Women in the top 6.9% of AI-assigned risk scores were invited to participate in the ScreenTrustMRI trial. A total of 559 participants completed supplemental MRI, resulting in the detection of 36 cancers, or 64.4 cancers per 1,000 MRI exams. The findings showed that AISmartDensity outperformed traditional breast density measures, with nearly four times higher cancer detection rates. Most cancers detected were invasive and smaller in size than those typically found in interval cancers. The majority of lesions were successfully biopsied, with some BI-RADS 3–5 findings requiring follow-up imaging. The cost of cancer detection was comparable to standard mammography, making the approach both efficient and scalable. Limitations include the need for further validation across different populations, the high upfront costs of MRI, and incomplete follow-up data for some flagged cases.