Case studies/Banking and financial services
A regional commercial bank re-keying loan files into nCino that a reader could have pulled from the signed term sheet on page one. We rebuilt the extraction path: LlamaParse on the term sheets and financials, LandingAI on signed PDFs, every field posted into nCino against the signed document with a field-level audit trail attached.
| File ID | Borrower | Amount | Agreement | Status |
|---|---|---|---|---|
| L-44021 | Ridgepoint Industrial | $2,400,000 | Term loan, 7 yr | posted |
| L-44022 | Maple City Holdings | $865,000 | Revolver, 364 day | posted |
| L-44023 | Northline Logistics | $3,150,000 | Equipment, 5 yr | pending |
| L-44024 | Cedar Mill Properties | $5,800,000 | CRE, 10 yr | review |
| L-44025 | Brightline Foods | $1,250,000 | SBA 7(a), 10 yr | posted |
| L-44026 | Harbor Marine Co. | $420,000 | Revolver, 364 day | posted |
| L-44027 | Foothill Health Group | $2,700,000 | Term loan, 5 yr | review |
| L-44028 | Westbrook Auto Parts | $610,000 | Equipment, 4 yr | pending |
At a glance
One commercial credit team, one LOS of record, one covenant library. The term-sheet-to-LOS mapping was the piece credit operations asked for first.
The engagement
The stack
ISO 27001 · ISO 9001 · GLBA scope · DPA and NDA signed at kickoff.
Before, the commercial credit desk
Commercial credit operations re-keyed the loan file. Every rate, every covenant, every collateral description was typed into nCino from a PDF already signed on the opposite side of the desk. These were the three patterns we found in discovery.
The term sheet carried the rate, the amortisation schedule, the covenants, and the collateral terms. Credit operations typed those 38 fields into nCino from the PDF. On a clean file, 35 minutes of processor time. On a file with a handwritten margin note or a scanned attachment, up to 70.
Pre-build baseline: 35 to 70 minutes of credit operations time per loan file, depending on cleanliness.
Covenant language on the term sheet was typed into nCino as a covenant record, which credit then monitored against. Every re-key was a chance for the covenant in nCino to drift from the signed version. Audit flagged the drift, but only sampled files caught it.
Pre-build baseline: measurable covenant drift between signed term sheet and nCino record, flagged in the audit sample.
When an audit team asked which document a figure came from, the processor had to open the loan folder, find the right page, and explain the extraction. That was fine at the file level. At scale across a portfolio review, it was the slowest part of the audit week.
Pre-build baseline: audit response time per sampled file ran to 20 minutes, per field.
What we built
The pipeline follows the same five stages we run on every loan file engagement. The covenant map and the field-level audit trail are the parts tuned against this bank's nCino structure.
Relationship manager mailbox polled on a 5-minute cadence, borrower portal webhook, branch scanner drop to SFTP. Every inbound assigned a single file ID before classification.
Document type tagged on ingest. Signed term sheet, borrower financial statements, collateral docs, covenant certificates, guarantor forms routed separately. Classification confidence below 0.92 holds the document for a processor tag.
Rate, amortisation, maturity, covenants, collateral description, guarantor. LlamaParse on typed term sheets, LandingAI on signed and scanned PDFs. Each extracted field linked to the page and bounding box it came from.
Every covenant field in the loan file checked against the signed term sheet. Financial statement figures cross-checked for internal consistency. Below 0.90 agreement, the file holds for credit operations review.
Clean loan files posted to nCino via the API. Each field carries a link back to the source document page and signature block. Exceptions routed to a named credit operations queue with the flag in plain English.
After, the numbers the credit desk signs off
Same credit staff, same relationship managers, same borrowers. The pipeline extracted the loan file against the signed term sheet and posted clean fields into nCino with the audit trail attached. Audit response time changed shape.
Credit operations still own the file. They still sign off every borrower and every covenant. The difference is that audit questions resolve to a page and a signature block, and the processors stopped re-keying 38 fields they never had to type.
From the desk
Audit used to ask where a figure came from and my team would open the folder. Now they click the field and see the signed page.
Commercial credit operations leadRegional bank, Carolinas
Handover
The engagement ends at a clean handover. The credit operations team runs the pipeline; Hexaa stays on call for a fixed retention period, then steps back.
Related cases
Each links to a named client, a named document, and the system the clean data lands in. We publish only what the client signed off to publish.
CIP completeness checked at intake, with a generated missing-items list routed back to the member before onboarding.
→Real estate · 2025CRE owner-operator · lease abstraction at acquisitionLease economic terms extracted into the abstract, with each abstracted field linked back to the page and clause it came from.
→Construction · 2025Commercial GC · submittal queue against the primeSubmittal log reconciliation against the prime contract spec. 1,200 submittals per project, Procore as the system of record.
→Free 30-minute call
You'll leave with a clear next step.
A loan file arrives with a signed term sheet, a borrower P&L, and a collateral description. The pipeline extracts 38 fields, posts them to nCino, and writes each field back to a page and a signature block. The audit question "where did this covenant come from?" resolves to one click.
