← All articles
AI & Technology

AI Across the Commercial Loan Life Cycle: Where It Does Real Work

Vijay Mehra
November 10, 2026

A relay race is rarely lost on the straightaways. It is lost at the handoffs, where the baton changes hands and half a second disappears.

A commercial loan has the same shape. The work inside each stage is usually sound. Time and accuracy leak where the file passes from the banker to the analyst, from the analyst to committee, and from closing to the people who watch the loan for the next five years.

Over nineteen months we talked with 90 lending institutions about how they underwrite and what they think of AI. Their answers map closely to those handoffs. What follows is a stage-by-stage look at where AI does real work today and what stays with people. Every figure describes the institutions we spoke with, and you can read the full study here.

First look

Thirteen institutions named pre-screen triage as the place deals stall. Seven shared enough of their funnel to compute a close rate, and the median was about 5%. At that rate, most analytical hours go to deals that never close.

AI changes the cost of the first read. An agent can read a partial package, reconcile what is there and flag what a credit committee would question, so the banker knows early whether the deal fits the box. Review that used to take weeks comes back the same day. The banker still decides whether to pursue it.

Underwriting

Twenty-five institutions described underwriting that lives in Excel and Word. Three put a number on the keying: 60% to 90% of analyst time goes to data entry. One credit union told us its analysts "spend 80, 90% of the time putting all the information in."

This is where AI earns its keep, on one condition: every figure has to be cited to the document and page it came from. Spreading, global cash flow and the reconciliation between documents are work a machine can do quickly. An analyst then confirms the figures that move a ratio or a policy test, and spends the saved hours on the borrower's story.

The credit decision

Thirteen institutions named memo writing as the bottleneck. Eleven asked whether AI can read a credit policy at all, and six said policy exception detection impressed them more than the underwriting.

A useful system checks each deal against the lender's own credit box, records every exception with its reason, shows the largest loan the deal supports and drafts the memo and a recommendation in the lender's template. Approval stays with the credit officers, and the record shows who approved what.

Closing

Closing is a checklist problem: conditions, documents and dates that have to match what was approved. AI can compare the closing documents to the approved terms and list what is still open. People clear the conditions and fund the loan.

After closing

Eighteen institutions named post-close monitoring as a blind spot. Covenant tests, borrower reports and key dates tend to live in a spreadsheet one person maintains.

Agents can start work when a borrower report arrives or a covenant test falls due, read the new financials and show how much headroom is left. They do not waive covenants or change risk ratings. A person reviews every result.

What stays with people at every stage

The questions lenders asked us most were about trust: 21 asked about data security, 20 asked why they should not use a general-purpose model, and 18 asked about accuracy. Those questions have the same answer at each stage. The lender's data stays isolated and is never used to train a model. Every figure can be traced to its source. A person approves each step.

LenderBox is built on those three rules, and it works beside the loan system you already have. If you want to see one of your own deals move through these stages, bring a deal.

Common questions

Where does AI help most in commercial lending?

At the handoffs: the first look at a new deal, spreading and reconciling the file, checking the deal against credit policy, drafting the memo and monitoring covenants after closing.

Does AI replace the underwriter?

No. The machine reads, reconciles and drafts. An underwriter reviews and approves at every stage, and credit officers make the decision.

What should a lender require from AI at every stage?

Three things: data that stays isolated and is never used to train a model, every figure traced to its source document and page, and a person approving each step.

Related reading

See it work on one of your own deals.

A 30-minute demo can run on documents you bring. Every figure comes back tied to its source page, and your team still makes every credit decision.

See it in action
30 MINUTES. NO COMMITMENT. 90-DAY MONEY-BACK GUARANTEE.