CRISPR interference (CRISPRi) is a versatile tool for high-throughput phenotypic screening. However, rational design and synthesis of the single-guide RNA (sgRNA) library required for each genome-wide CRISPRi application is time-consuming, expensive, and unfeasible if the target organisms lack comprehensive sequencing and characterization. We developed an ultra-dense random sgRNA library generation method applicable to any organism, including those that are not well-characterized. Our method converts transcriptome-wide mRNA into 20 nt of sgRNA spacer sequences through enzymatic reactions. The generated sgRNA library selectively binds to the non-template strand of the coding sequence, leading to more efficient repression compared to binding the template strand. We then generated a genome-scale library for Escherichia coli by applying this method and identified essential and auxotrophic genes through phenotypic screening. Furthermore, we tuned the production levels of lycopene and violacein and identified new repression targets for violacein production. Our results demonstrated that a genome-scale sgRNA library can be generated without rational design and can be utilized simultaneously in a range of phenotypic screenings.