FAQ
AI Questions
Scroll
Is Sweet's AI decision-making fully autonomous?
No. A human review step is always in place by default. The AI is most powerful as an efficiency layer: it checks documents for completeness and validity, flags errors and inconsistencies, extracts and validates data, and applies pre-configured criteria to drive automated decisioning and pricing. The loan officer stays in control throughout origination.
Link to this answer
What's "under the hood" of the AI? Whose models, and does it handle handwritten documents?
OCR plus LLMs (not self-trained), running entirely inside your own secure environment. The lending institution defines what to check and extract; it works on any document format, including handwritten ones, and won't fabricate. It explains what it found or didn't.
Link to this answer
How does Sweet's AI improve borrower experience without removing the lender relationship?
It simplifies the borrower's tasks, gives clearer guidance, and reduces confusion while keeping the lender in control of communication, review, and final decisions. Borrowers get a more modern, guided experience, but the relationship stays centered on the lender's expertise and judgment.
Link to this answer
How does the AI product help us place loans we otherwise might decline?
It can screen the loan against the funding routes open to it and identify outside funding sources or capital partners whose buy-box may fit, helping lenders preserve the relationship instead of turning the borrower away. This gives lenders flexibility to support borrowers even when a loan doesn't fit their own balance sheet, concentration limits, or internal credit appetite.
Link to this answer
Can AI reduce the back-and-forth between borrowers, lenders, CPAs, and third-party funding sources?
Yes. AI can organize requests, flag missing items, review documents, extract data, and keep each party aligned on what still needs to be done. Instead of chasing updates through email threads, each stakeholder sees the right tasks, documents, comments, and next steps in one controlled workflow.
Link to this answer
Can the platform automatically request different documents depending on borrower answers?
Yes. Sweet can trigger conditional document requests based on loan type, borrower structure, entity role, loan amount, collateral, or other configured criteria, helping lenders collect the right information the first time without overwhelming borrowers.
Link to this answer
Can we see when a human approved, rejected, or changed an AI-generated result?
Yes. Human review actions can be tracked, so teams can see who approved, rejected, edited, or overrode an AI-generated result. This keeps AI-assisted work auditable and gives credit, compliance, and operations teams confidence that final decisions remain under human control.
Link to this answer
Can the AI replicate the structure of our existing underwriting workbook?
Yes. Sweet can use your existing underwriting model structure to extract, spread, and organize data in a format your team already understands, so lenders benefit from automation without abandoning the credit models, formulas, and review habits they trust.
Link to this answer
Can Sweet AI flag inconsistencies between the application, tax returns, balance sheet, collateral schedule, and other documents?
Yes. The AI compares information across documents and flags mismatches, missing details, outdated information, or values that don't align. This helps underwriters spot issues earlier, reduce manual review time, and improve the quality of the loan package before a decision is made.
Link to this answer
See the platform on your own book of business.
A working demo with your loan programs, your documents, and your workflow, not a slide deck.