A proposed system for moving a large asset-based loan application process from paper to a fully agent-driven workflow was designed to work without needing clean or standardized data, as generative AI can read the paper trail as it exists. The system could route application pieces to specialized functions and flag items for human review, with one agent able to read a financial statement and enter figures directly in milliseconds, including a human check that sampled output regularly. The bank’s actual process involved manual data entry and re-uploading into a second system, taking several hours with no equivalent check. The proposed process could have gone from weeks to days, but the bank instead applied older technology to a couple of steps, dropping the timeline by a couple of days with savings close to nothing. The bank hesitated because of trust, not technology. A 2026 Grant Thornton survey found that half of banking executives say governance and compliance concerns are already limiting how well their AI performs, and only 18% feel confident their AI controls could pass an independent audit.