Innovating PPP Loan Processing for CRFG

When the Paycheck Protection Program launched, CRFG was flooded with loan applications almost overnight. Their existing loan processing workflow was designed for a fraction of the volume they were now facing. Each application required extracting data from submitted documents (tax returns, payroll records, bank statements), validating the information against program rules, and preparing the loan package for SBA submission. Staff were working around the clock, manually entering data and cross-checking figures. Applications were piling up, and borrowers who desperately needed funds were waiting weeks for processing. CRFG needed to dramatically accelerate throughput without sacrificing accuracy, because SBA compliance requirements meant that errors could result in rejected applications or clawbacks.

We built an automated document processing pipeline that could handle the full intake-to-submission workflow. The system uses OCR and intelligent document parsing to extract key fields from uploaded documents (payroll figures, tax IDs, employee counts, loan amounts). Extracted data is automatically validated against PPP program rules, with the system flagging applications that have discrepancies or missing information for human review. Clean applications move through the pipeline to package preparation and SBA submission with minimal manual intervention. We built the system to handle the specific document formats that PPP applications required, and trained extraction models on real application packages to achieve high accuracy. A review dashboard gave CRFG staff visibility into the pipeline, letting them quickly address flagged applications and track overall throughput. The system went from concept to production in under three weeks to meet the urgency of the program timeline.