Example processfinanceTRID/ECOA

Mortgage underwriting automation software

An example of mortgage underwriting as a governed process: agents do the 9 steps, and a person signs the gate before anything irreversible happens. Download it, import it into your workspace, and adapt it to your systems.

Download example · JSONImport into Minctrl →

minctrl.process/v1 · hash f8dc2b3e09d5…d40c

The graph9 agent steps1 human gatesigns: DE-authorized underwriter

Mortgage-underwriting autopilot

It costs ~$9,000 to originate a loan — mostly processor and underwriter labor against codified guidelines — and any model-driven decision invites fair-lending scrutiny. Documents are classified, income/assets verified, and eligibility decided against agency guidelines straight-through; the DE underwriter signs the approval and clear-to-close.

The governed flow · branches, parallel work & a human on the irreversible step
Approve / Eligible (in policy)Refer / Refer-with-CautionMissing / stale docs — reworkAll conditions cleared
Start
agent · low risk
Intake the application (1003) and borrower documents
parallel · fan-out / join
Run verifications concurrently
agent · low risk
Verify income and assets against agency guidelines
agent · low risk
Pull credit and tradelines from the bureaus
agent · medium risk
Run automated underwriting (DU / LPA)
decision
AUS recommendation?
agent · low risk
Auto-build the in-policy approve/eligible file from AUS findings
agent · medium risk
Manually underwrite the referred file and document compensating factors
agent · low risk
Apply investor overlays, run ECOA / fair-lending + TRID checks, and build the condition list
decision
Conditions cleared?
gate · human sign-off
A DE-authorized underwriter signs the approval and clear-to-closesigns: DE-authorized underwriter
agent · high riskirreversible
Issue the approval and clear the loan to close
agent · low risk
Monitor conditions, early-payment-default and HMDA reporting
Done
low riskmedium riskhigh riskbranch / reworkgate · human sign-off
What it is modelled on

A denial is an ECOA adverse action with specific-reason and notice obligations, so an underwriter signs off before an adverse decision (or an exception approval) is issued.

RuleWhat it requires
TRID (TILA-RESPA Integrated Disclosure)Borrowers must get a Loan Estimate within 3 business days of application and a Closing Disclosure at least 3 business days before consummation; certain changes re-start the timing clock.
ECOA / Regulation BCredit decisions may not discriminate on prohibited bases, and an adverse action requires a notice with specific principal reasons within 30 days.
Ability-to-Repay / QM (Reg Z)The lender must make a reasonable, good-faith determination of the borrower's ability to repay using verified income, assets and obligations.
FCRA adverse-actionIf a decision is based on a credit report, the adverse-action notice must include the FCRA-required disclosures.

A summary for orientation, not legal advice.

Use this example
01

Download the file

A minctrl.process/v1 document: the graph and the hash of the process it becomes.

02

Import it in Processes

Import into Minctrl opens this example in your workspace; one click stores it as your first version. Import JSON does the same with the file, and a file changed after download is refused.

03

Connect your systems

Every connector starts in shadow: steps run, nothing leaves Minctrl until you configure a connector and switch it live.

04

Adapt it

Change signers, time limits and steps on the canvas. Each save is a new version; a running case stays on the version it started with.

Questions
How does the mortgage-underwriting example handle TRID and ECOA timing?

The example models the clocks: the Loan Estimate within 3 business days of application, the Closing Disclosure at least 3 business days before consummation, and ECOA/Reg B adverse-action notices with specific reasons within 30 days. The timeline is recorded as evidence. Whether your disclosures and notices comply depends on your own documents, systems and legal review. Minctrl holds no certification, and this is not legal advice.

How does the AI make a fair-lending-safe decision?

The AI agent verifies income, assets and obligations and runs Ability-to-Repay/QM and policy checks, producing a recommendation. Denials and exception approvals park at an underwriter sign-off gate, because an ECOA adverse action carries specific-reason and notice obligations that a human owns.

What evidence do we have if a fair-lending exam questions a denial?

Every underwriting decision is logged with the verified inputs, the policy applied, the principal reasons, the underwriter who signed off, and a tamper-evident audit hash. Deterministic, replayable runs let an examiner reconstruct exactly why an application was approved, conditioned, or denied.

More examples