Skip to content

AI Underwriting: automated, confidence-scored credit memos

Underwriting (sidebar → Workflows → Underwriting) turns a folder of deal documents into a structured credit memo in about a minute. Upload PDFs, spreadsheets, and decks; five AI agents extract the company profile, financials, collateral, and risk factors; and the result is a memo where every extracted field carries its own confidence score — exportable to DOCX and Excel (plus CRE decks).

Wagoo AI underwriting deals list with pipeline stats, confidence scores, and per-deal status

Four stat cards summarize the book: Total Deals, In Review, Total Pipeline (sum of facility amounts, excluding failed deals), and Completed.

The table below lists every underwriting deal with its Company, Status, Confidence (percent + mini bar, color-coded), Sector, Files, and Date. A row menu (⋮) offers View and Delete (confirm “Are you sure you want to delete this deal?”). Empty state: “No deals yet / Upload your first deal to get started.”

  • Filter by status with the All Deals dropdown: Draft, Processing, Review, Approved, Rejected, Failed.
  • Toggle between table and card grid with the view switcher (remembered; mobile forces the card grid).
  • New Deal starts an upload.

Each deal carries a product-type badge on its detail header — Asset-Based Lending, CRE Bridge, Equipment Finance, or C&I Term Loan — which determines the memo sections. (A “Residential” type exists in the system but isn’t customer-facing yet.)


Click “New Deal” to open the upload view.

The Wagoo underwriting upload view with product type selector and drag-and-drop document dropzone

  1. Product Type — leave “Auto (enterprise default)”, or pick one. The customer-facing choices are Asset-Based Lending, CRE Bridge, and Equipment Finance. (This selector only appears when your workspace is enabled for multiple product types.)
  2. Add documents — the dropzone reads “Upload Deal Documents / Drag and drop files or click to browse”, or use “Select Files”. Supported: PDF, Excel (.xlsx, .xls, .xlsm), Word (.docx), PowerPoint (.pptx), up to 50MB per file (duplicate filenames are deduped). Typical inputs: financial statements, borrowing-base certificates, AR/AP agings, rent rolls, appraisals, equipment schedules, pitch decks.
  3. Click “Process N File(s)”.

More documents generally mean better extraction and higher confidence scores — the agents cross-reference sources. You can add files to a deal later and reprocess.


The processing view shows a live step name, details, and a progress bar. Under the hood:

  1. Uploading — your files are stored against the new deal.
  2. Extracting Text — every document is parsed (including scanned PDFs, which go through vision transcription).
  3. AI Analysis — three extraction agents run in parallel: the company agent (profile, ownership, sector), the financial agent (income statement, balance sheet, spreads), and the collateral/field-exam agent (AR/AP/inventory, equipment, or property, by product type).
  4. Enrichment & Risk Analysis — two more agents run in parallel: web enrichment (public information about the company) and the risk agent (risk factors and mitigants).
  5. Building Deal Terms → Finalizing Analysis — results are cross-checked, confidence-scored per field, and assembled into the memo.

The visible step names are Uploading files → Starting Analysis → Extracting Text → Text Extraction Complete → AI Analysis → Enrichment & Risk Analysis → Building Deal Terms → Finalizing Analysis. You can Cancel while it runs, and “Back to Deals” on a failure or timeout. Long jobs keep running in the background — if the screen times out with “Still processing,” the analysis continues (polling gives up after ~60 minutes) and the deal appears in the list when done.


When processing completes, the deal opens on its memo.

Top of a Wagoo underwriting deal: status, confidence score, export buttons, and company information

The header shows the company name, status pill, deal ID (click to copy a shareable URL), product-type badge, upload date, and the overall Confidence score, colored by strength.

Below the header, the memo is a stack of sections whose exact card titles and mix depend on the product type:

Section (card title)Appears forContents
Company InformationAllName, sector, location, year established, NAICS/SIC
Ownership & ManagementAllOwners, stakes, key people
Deal TermsABLFacility size, advance rates, pricing
Collateral & Borrowing BaseABLAR/inventory eligibility, reserves, availability
Property & Valuation (+ Sources & Uses, Equity Returns, Operating Pro Forma)CREValue, NOI, cap rate, comps, and the CRE financial modules that stand in for corporate financials
Equipment & Structure, Credit Scorecard, Financial Spread, PayNet AnalysisEquipment FinanceAsset schedule and credit scorecard
Leverage & CovenantsC&I Term LoanLeverage ratios and covenant analysis
A/R Concentration (Top 10) & A/P Concentration (Top 10)All (when data exists)Top-10 customer/vendor concentrations
FinancialsNon-CRERevenue, EBITDA, margins, trends
Balance SheetABLAssets, liabilities, working capital
Risk AssessmentAllKey Considerations, Strengths, Risk Factors, Mitigants, Exit Strategy
Industry & PositioningAllMarket context from web enrichment

An asset-based lending borrowing-base section in a Wagoo underwriting memo, showing availability and additional collateral

Every field is confidence-scored (0–100, color-coded) so you can see at a glance which numbers came through clean and which need a human look. Fields the documents didn’t support show “Add value” instead of a guess, and a per-card “Show more fields” expander lets you fill in catalog fields the extraction left empty.


Editing fields, provenance, and reprocessing

Section titled “Editing fields, provenance, and reprocessing”

The memo is a working document, not a printout:

  • Edit fields inline — corrections save to the deal and the memo re-derives immediately, no reload. (On deals backed by extraction claims, editability is per-field; derived/computed fields aren’t directly editable.)
  • Revert — a field you’ve overridden shows a “Revert to extracted value” control (toast “Reverted to extracted value”).
  • Provenance — the ⓘ next to a value shows its source line, validation, and origin document (with cell/page), or “No source document on file yet.”
  • Concurrent edits — if someone else changed the deal, a banner appears: “This deal changed since you loaded it. Refresh to get the latest values.”
  • Deal Documents — a collapsible card (default collapsed) listing every uploaded file with in-app preview and download, plus an add-files button.
  • Reprocess — re-runs the full analysis over every attached document. It’s available from the Deal Documents card only when the deal’s status is Review or Failed, and confirms first: “All current analysis will be replaced. The pipeline will re-run on every document currently attached to this deal… Manual edits to extracted fields will be overwritten.” Reprocess after adding significant new documents; edit inline for small corrections.

Export buttons sit at the top of every completed deal:

ExportFormatAvailable
DOCXFull written credit memoAll product types
ExcelStructured workbook of the extracted dataAll product types
PresentationPDF deckCRE deals
4-PagerCondensed DOCX summaryCRE deals

Exports are disabled while a deal is processing or failed.


Use the status buttons in the header to record the credit decision:

StatusMeaning
DraftCreated, analysis not final
ProcessingAI analysis running
ReviewAnalysis complete, awaiting decision
ApprovedAccepted
RejectedDeclined
FailedAnalysis errored (see the validation message; fix inputs and reprocess)

Approved and Rejected deals show a “Reopen” button to bring them back into Review. Draft and Review are disabled while a deal is processing; Approve and Reject are disabled while processing or failed.


How underwriting connects to origination and the CRM

Section titled “How underwriting connects to origination and the CRM”
  • From origination: promote an application with “Send to Underwriting” — Wagoo creates the underwriting deal, starts analysis immediately, and freezes the application as an audit snapshot that links to the deal.
  • From the assistant: the AI Assistant can look up and read your underwriting deals in chat (“what did the analysis say about Acme’s concentration risk?”).
  • Direct upload works too — underwriting doesn’t require origination or the CRM.

How long does analysis take? Usually about a minute; large document sets take longer. Analysis continues in the background even if you navigate away or the progress screen times out.

What documents should I upload? Whatever you’d hand a human analyst: financial statements, agings, borrowing-base certificates, rent rolls, appraisals, tax returns, equipment schedules. PDF, Excel, Word, and PowerPoint are supported, 50MB per file. Scanned documents are fine — they’re transcribed automatically.

What does the confidence score mean? It’s the system’s own estimate of extraction reliability, per field and overall. High-confidence fields were corroborated by the documents; low-confidence fields deserve a manual check. Editing a field yourself makes it authoritative.

Will reprocessing lose my edits? Yes — reprocess replaces the entire analysis, including manual field edits (the confirmation dialog warns you). Export first, or prefer inline edits when the change is small.

Can I re-run just one section? No — reprocessing is whole-deal. Inline edits are the way to adjust individual fields.

Who can see underwriting deals? Everyone in your workspace with the underwriting feature. Deals are isolated per workspace.


Related: Loan Origination · AI Assistant · Reference: statuses