Use caseBusiness lending
Whitaker Millwork wants $1.8 million for a new line. Someone keys in three years of financials first, and the memo waits for the last night.
The key figures are read from the financials with their pageTwo points are flagged before the analysis startsThe analyst signs the memo, the committee decides
Points need the analyst · the AI names each one, with its page
All required documents are on file. One point for the analyst: the notes list $420,000 of lease commitments kept outside the balance sheet.
source · 2025 financial statements, note 14What does a credit analyst do on Monday with a $1.8 million request? She keys in three years of statements and checks the totals twice. The lease note turns up on Thursday, nobody sets the plan's start date against the supplier's lead time, and the memo is written in one evening. By the time the analysis starts, half the week is gone.
Here the reading happens before Sarah arrives. The capture chain reads the key figures with their pages, and the AI checks the file against your credit policy and the documents against each other. Sarah spends the week on the restatement, the covenants and the recommendation, and the committee decides on a memo where every figure opens its source.
Spreading, policy checks, memos: every credit desk knows these screens. Here they come in a new order: the figures are read with their pages, the two traps are flagged on day three, the analyst tests and recommends, and the committee decides on a memo it can check. The nine steps below run from Andre Wallace opening the request on November 2 to the drawdown on December 3.
The banker opens the request and the company sends its documents. They are read before the analyst picks the file up.
Andre Wallace, the business banker, meets the management of Whitaker Millwork Inc., a millwork manufacturer near Raleigh, North Carolina, with $9.4 million in revenue. They want $1.8 million over seven years for an automated production line. He opens a business loan case and attaches the equipment quote.
The case lists one slot per document your credit policy asks for at this amount. Isabel Whitaker, the CFO, uploads the rest from the business portal.
The capture chain classifies each document, identifies its year, and extracts the values your credit team defined for a business file: revenue, EBITDA, operating income, net financial debt, equity, lease payments. Each value is linked to the page and line it was read on.
Sarah Klein, the credit analyst, approves or corrects each value on the review screen. Approved values are written to the case.
An AI agent runs the file against the rules your credit team wrote in plain language, and checks the documents against each other.
The status is Attention required, with two points. The analysis writes the reasoning on the case and cites each page.
The historical figures hold. What stays open is whether the company can carry the loan before the line produces.
Sarah opens note 14 of the 2025 financials at the cited page. Your credit policy treats leases as debt, so she restates $420,000 of commitments into net financial debt and records the adjustment with its reason. The case keeps the value as read and the value as adjusted, side by side.
Sarah asks the assistant what the 2027 figures rely on. The answer quotes the plan: the line runs from April, adds $900,000 of revenue in its first year, and keeps the gross margin flat. It sets the April date beside the quote's lead time.
Sarah runs the covenants your policy sets for an equipment loan in your credit model, with the restated debt: net debt to EBITDA at most 3.0x, debt service coverage at least 1.25x. Coverage falls to 1.15x in 2027. With a twelve-month interest-only period, it reaches 1.8x. She files the test on the case.
Sarah has her recommendation. The memo starts as a sourced draft and leaves her desk with her name on it.
Sarah runs Draft the credit memo, a business task set up in the console with your committee format. The draft takes the approved figures, her adjustments, her covenant test and the two flagged points. Every figure links to its source.
She rewrites the risk section in her own words, adds her recommendation and signs. Before submission, the AI checks the memo's covenants against your policy for this loan type.
On November 19, the committee opens the file from its agenda. Members read the memo, open the plan and the covenant test in the viewer, and ask Sarah about the installation date. They approve with the interest-only period, the lien and one added condition: the CEO's personal guarantee for $300,000.
The decision opens the follow-up tasks: draft the commitment letter from the committee conditions, file the signed letter, collect the guarantee and the lien before drawdown, then test the covenants each May on the new financials, with an SLA timer on each.
Whitaker's file raised two flags. Against your credit policy, every loan file ends up one of three ways, and each one hands the analyst a different next move.
Documents on file, figures consistent, plan and quotes aligned. The analysis notes each rule with the page that settled it.
What follows: the analyst reviews the figures, tests the covenants and recommends.A lease outside the balance sheet, a date that does not fit. Each point names its pages.
What follows: the analyst adjusts or asks the borrower, with the reason recorded.A sector or a structure your lending policy excludes. The AI lists what passes, then names the rule.
What follows: the head of credit reviews and decides whether the file goes further.It lends to nobody. It declines nobody. It computes no rating and sets no condition. Across the nine steps on this page, the restatement, the covenant test and the recommendation are the analyst's, and the decision is the committee's.
The AI reads, checks, compares and drafts. The control is in your credit policy and in the committee's approval path, both written into the process.
Three counts from the Whitaker request. A borrower with clean notes and a firm delivery date would raise fewer flags.
Where these numbers come from. They are counted on the scenario described on this page, with example data. They are not a customer result.
The analyst's desk changes first, but the head of credit and the architect have their say too. This is what each of the three finds in it.
Keying figures takes the week. The memo gets one evening.
In the meeting, a figure you cannot trace is a figure you take on trust.
Today the loan file is split across a portal, a shared drive and a credit model.
Choose a check your analysts do by hand on every file, such as leases disclosed in the notes, and run it on arrival. A month of loan requests will tell you whether it earns its place.
Your business portal, your credit model and your case files keep doing their jobs. The reading and the checks sit on top.
Next step
Bring a few anonymized business loan files and your credit policy checklist. We will read them with you and show where each figure lands in the memo.