Mortgage QC automation: every closed loan reviewed before it is sold, not a sample after
Every loan reviewed before it is sold, not a sample after.
See one case, screen by screen ↓Software that runs a lender’s quality control review on closed loans automatically. Mortgage QC Automation reads every closed loan file the moment its closing package is complete, runs a 386-check QC script, recalculates income, debts, assets and fees, checks every disclosure and its tolerances, and grades each defect by agency severity — before the loan is sold.
How to choose AI solutions for lending →
Updated
The agency rules ask for a monthly post-closing sample — 10% random or a statistically valid sample — reviewed within 90 days of the end of the closing month. Most loans are pooled and sold well before then. So the QC review usually finishes after the sale, and a defect found then can only be cured at a cost, repriced or repurchased.
Inside the sample, the work is still by hand: auditors sort and bookmark the file, re-key the numbers, recalculate income and DTI, line up the loan estimate against the closing disclosure, and wait on VOEs, credit refreshes and tax transcripts. The other nine loans in ten get no post-closing review. Defect rates swing with the refinance mix, and a QC team cannot grow fast enough to keep up.
Estimated split for one loan in the post-closing sample, by hand and with the solution, in days from a complete closing package.
A 386-check QC script runs on every closed loan before it is sold; agents grade each defect by agency severity and QC auditors confirm it.
The Dorsey purchase is the last of Halden Ridge Home Loans’ 1,186 September closings. Its closing package was complete on Fri 9 Oct; its pool cuts off Fri 16 Oct. Here is what happens, screen by screen, in the working solution.
Renée opens the solution on the month as it stands: 1,185 of 1,186 September loans run through the full QC script before sale, 13 held from the October pools. Each square is a loan, coloured by its grade. Pressing “10% sample” dims every loan a sample would never have read.
“A 10% sample: 119 loans, reviewed by about day 75, after the pools settle.”
Marcus and Elena Dorsey, loan 2609-04417: a $412,800 2-unit purchase, closed Thu 24 Sep at the Westlake branch, slotted for the pool settling 21 Oct. The script runs 386 checks in nine parts, each headed by the agent that runs it. Every row shows the value read and the page it came from — note, deed, credit report, closing disclosure.
Qualifying ratios, check RAT-005: DTI recalculated at 51.2% against 44.8% approved, over the 50% maximum. Qualifying income falls from $13,370 to $12,770; monthly debts rise from $5,989.33 to $6,538.23. The defect grader marks it significant — not eligible for sale as closed.
“Agency severity: the loan was not eligible for sale as closed. Curable: document the student loan payment and re-run the AUS — at its $186 income-driven payment, DTI is 48.4%.”
Each cause links to its source. The lease for unit B shows rent of $2,400 a month, counted in full on the transmittal; the guide counts 75% of gross rent (B3-3.1-08). The credit report shows Elena’s deferred student loan of $54,890 counted at $0, where the guide counts 1% of the balance (B3-6-05).
Every failing check opens the document page that caused it.
Renée continues the script to the end: 386 checks, 375 passed, 9 failed, 2 need review. Four findings, each with its agency grade — 2 significant, 1 moderate, 1 minor. The loan is held from the 21 Oct pool, with both significant findings curable before the Fri 16 Oct cut-off.
The loan review shows the six agents’ run step by step and every finding with its links. Two are significant: the DTI defect, and an $18,500 gift from Elena’s father with no proof of transfer, which leaves the file $8,060 short without it. One is moderate: a $175 re-inspection fee with no valid change of circumstance and recording fees $38 over the 10% tolerance — a $213 refund. The fourth, minor: the initial application unsigned by the loan originator.
The agents propose. Owen confirms or rebuts each finding, and a rebuttal goes back to the agent with its reason.
With all four findings confirmed, Renée signs and sends the cure plan: correct the rental income and re-run the AUS by Wed 14 Oct, refund $213 with a letter and corrected closing disclosure by Thu 15 Oct, and the donor’s statement for the gift by Thu 15 Oct. Tasks go to Caleb Ruiz, Mei-Lin Ostrander and the loan officer, Brett Sayer.
“Done by Thu 15 Oct, the loan makes the 21 Oct pool — one day before the cut-off.”
The delivery gate shows each October pool with its loans in and held: 14 loans held, $4.68M of $398.6M — 11 curable before the cut-off, 3 not curable (2 to the portfolio desk, 1 to fraud review). No loan with an open significant finding goes into a pool. The hold list goes to Jordan in secondary with one click.
Trends put the Dorsey defect in context: Westlake runs at 24.3 findings per 100 loans against 17.9 for the company, led by rental income on 2–4 unit homes. The fix is a corrective action — training for Westlake underwriters, and a warning at underwriting when rent is counted at 100% — owned by Caleb, due Fri 30 Oct.
The trend and report writer drafts the September report for senior management: 1,186 loans reviewed, a significant defect rate of 1.18% against the industry’s 1.71%, and 0.25% sold with a significant defect after cures, down from 1.52% in May. Nothing is sent until Renée signs it for Denise Hollander, SVP of quality and risk.
The same loan, seen by the four people who carry it — QC, underwriting and secondary — and what their month looks like now.
A 386-check QC script runs on every closed loan before it is sold; agents grade each defect by agency severity and QC auditors confirm it.
Classifies every document in the closing package, extracts its data and compares it with the loan data in the origination system; checks that required documents are present and signed.
Recomputes qualifying income, debts, DTI, assets, gifts and large deposits from the documents — salary, overtime and bonus with history, self-employed income, rental income at 75% of gross rent.
Lines up every loan estimate and closing disclosure, tests zero and 10% tolerances, checks each change of circumstance, disclosure timing, APR tolerance, right to cancel for every owner and HMDA data, and computes any cure and its deadline.
Orders and reads re-verifications for the samples: employment, a refreshed credit report, tax transcripts and an appraisal desk review.
Maps each finding to the defect categories and agency severity, decides whether it can be cured and how, with the owner and deadline, and holds significant loans from delivery.
Tracks defect rates by branch, product, underwriter and category, finds root causes and drafts the monthly QC report.
Confirm each finding; release cured loans; sign the report. The agents propose; a named person decides.
The QC team can ask in plain words — or ask for a draft. Every answer points to its source.
Why is the Dorsey loan held from the 21 October pool?
Two significant findings and one moderate, all curable before the 16 October cut-off. Significant: rent counted at 100% of the $2,400 lease, not 75%, and a deferred student loan at $0 — DTI 51.2% against 44.8% approved. Significant: an $18,500 gift with no proof of transfer — $8,060 short. Moderate: a $175 re-inspection fee with no valid change and recording fees $38 over 10% — a $213 refund. Cured by Thursday 15 October, it goes back into the pool.
Which branch is driving our defects this quarter?
Westlake: 24.3 findings per 100 loans in September against 17.9 for the company, with 3 of the 14 significant loans. The pattern is rental income on 2–4 unit homes counted in full — 5 of the quarter’s 7 such findings, 4 from one underwriter. A training action for Caleb Ruiz is already in the September report.
How does our significant defect rate compare with the industry?
September: 1.18% of loans had a significant defect; the industry’s critical defect rate — the same grade under its name — was 1.71% in Q1 2026, up from 1.38% in Q4 2025. And 0.25% of September loans will be sold with a significant defect after cures, against 1.52% in May, when only the 10% sample was reviewed, after sale.
Draft the fee refund letter to the Dorseys
Drafted for Mei-Lin Ostrander to sign. It explains the $175 re-inspection fee and the $38 over the recording-fee allowance, and encloses a $213 refund and a corrected closing disclosure. The legal deadline is 23 November, 60 days after closing; the cure plan has it by Thursday 15 October so the loan makes its pool.
13 screens from the working solution, on its sample data. Pick one to see it large.
The rules every loan is reviewed against are the lender’s own — versioned, approved and on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Review scope | Every closed loan, before delivery | On or off |
| Agency random sample | 10% of closings, monthly | 10% monthly · statistical, 95% confidence, 2% precision |
| Discretionary samples | Early defaults, DTI over 45%, 2–4 units, gifts, new loan officers | Early defaults and DTI over 45% only · every manually underwritten loan |
| Moderate defects | Cure before sale | Cure before sale · fix within 30 days · blocks delivery |
| Pool cut-off | 3 business days before settlement | 2 · 3 · 5 business days |
| Rental income for qualifying | 75% of gross rent (lease or rent schedule) | 75% of gross rent · net of expenses from the tax return |
| Verification of employment | Within 10 business days of the note | Within 10 · within 5 business days |
| Who releases a held loan | Renée Castellano, QC manager | Renée Castellano · Owen Mbeki |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample loan data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (ACES Quality Management mortgage QC industry trends, Q1 2026, via National Mortgage News, 18 Aug 2026 — critical defect rate 1.71%; Fannie Mae Selling Guide D1-3 — 10% or statistical sample, reviewed within 90 days). Halden Ridge Home Loans, its branches, borrowers and the people named on this page are fictional — characters and sample data in the working solution.
Software that runs a lender’s quality control review on closed loans automatically. Mortgage QC Automation reads every closed loan file the moment its closing package is complete, runs a 386-check QC script, recalculates income, debts, assets and fees, checks every disclosure and its tolerances, and grades each defect by agency severity — before the loan is sold.
No. The 10% random sample and the discretionary samples are still drawn and re-verified in full by people. In this solution the random sample also measures the agents: in the working solution, auditors confirmed 97.8% of agent findings on it over three months.
By agency severity, set in the QC plan. Significant means not eligible for sale as closed and blocks delivery; moderate is eligible but must be cured before sale; minor has no effect on eligibility and is counted in the trends. FHA loans map to the HUD defect tiers.
Yes. A loan with an open significant finding is held from its pool. The delivery gate shows each pool’s held loans, the cure, its owner and due date against the cut-off, and the QC manager releases a loan once its cure is confirmed.
The disclosure and compliance checker lines up every loan estimate and closing disclosure, tests zero and 10% tolerances, checks each change of circumstance and its timing, and computes the cure and its deadline — 60 days after closing for a tolerance refund. It drafts the cure; the compliance officer sends it.
Yes. The agents propose findings; a QC auditor confirms or rebuts each one with a reason. The agents never change loan data or an underwriting decision, and never contact a borrower. The QC manager signs the cure plan, releases held loans and signs the monthly report.
The closed loan file and closing package, loan data from your origination system, re-verification sources (employment, credit bureau, tax transcripts, appraisal desk review), the selling guides and your QC plan, and your pool calendar for the hold list.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your QC plan, defect categories and a sample of closed loan files — and it goes live once every quality gate has passed. We will walk you through it on your own loans first.
An income worksheet in minutes, every figure traced to the return.
1-min demoMortgage Condition ManagementConditions cleared the day the document arrives, not days later.
1-min demoMortgage Guideline SearchA cited guideline answer in seconds, plus weekly change notes.
1-min demoWe’ll run Mortgage QC Automation on a sample of your own closed loan files.