Mortgage lending

Let the underwriting agent clear the clean files

Clean applications move toward closing without waiting for a person. A file that breaks a lender rule stops, and a close call goes to an underwriter with every number the agent used.

See an agent at workTry the loan simulator

Connecting the agent? See the developer guide for Mortgage & Loan Processing

Live demo

Try a loan decision

Choose a borrower, run the decision, and review the governing rule.

Demo is live

How the simulation works

The website proposes approval. VeriProof checks the rules.

  1. The website simulates the agent

    Every prepared scenario starts with the website acting as an AI agent and proposing approval.

  2. VeriProof enforces the business rules

    VeriProof checks the proposed approval against the rules prepared for this demo.

  3. The website applies the result

    VeriProof returns one of five outcomes: Allow, Warn, Transform, Escalate, or Deny. The website applies that outcome to the simulated action.

In a real integration, your application applies VeriProof's result before the proposed action proceeds.

  1. Choose what the agent approvesEvery prepared scenario starts with simulated agent approval
  2. See VeriProof's decisionPass, escalate, or stop based on the business rules

Choose a scenario

The loan application

Each borrower is set up to end differently. Pick one, or change any detail below.

Loan purpose
Home purchase
Occupancy
Primary residence
Property type
Single-family home
Home price
$400,000
Down payment
12%
Change scenario details

This adds up to a $352,000 loan covering 88% of the home price, with debts taking 13.5% of monthly income.

The app proposes approval. VeriProof stops it because Taylor's credit score is below the lender's minimum of 680.

Records created in this public demo can be seen by other visitors, so use the sample data rather than your own.

Configuration details

How this demo is set up

Review the application settings, rules, and evidence fields used in this simulator.

VeriProof setup

Industry
Mortgage & Loan Processing
Application
Automated Underwriting Engine
Environment
Public demo workspace
Rule enforcement
On, rules can stop actions
Rule set
Mortgage underwriting guardrails

Configured frameworks

NIST AI RMFOSFI E-23

The app's own checks

The sample app proposes an outcome from the loan facts. These thresholds belong to the demo, not a lending standard.

  • Propose denial when the credit score is below 620 or monthly debts exceed half of monthly income.
  • Send the loan to a person when it is above $650,000 or covers more than 97 percent of the home price.
  • Propose approval when the score is at least 640, debt share is at most 45%, and the other limits are met.
  • Send remaining applications for underwriter review.

The company rules VeriProof enforces

VeriProof checks the proposed action against the lender’s rules. A stop or review requirement changes what can proceed.

  • Stop any approval where the credit score is below 680.
  • Send any approval to an underwriter when monthly debts are above 43 percent of income.
  • Require underwriter review for loan-to-value ratios of 95% or more unless the session already records that escalation.
  • Permit the proposed approval when no rule stops it or requires review.
Review the details saved with each run

A saved run connects the loan facts, app proposal, and policy result. Check its save status to confirm the record is available.

applicant_name

The borrower’s name, kept with the record so a reviewer knows whose loan this is.

applicant_id

A stable borrower ID that makes the record easy to find and link to later.

loan_purpose

Whether this is a purchase, a refinance, or a cash-out loan.

occupancy_type

Whether the borrower will live in the home, use it as a second home, or rent it out.

property_type

The kind of property backing the loan, house, condo, or multi-family building.

credit_score

The score both the app and the company rules used to decide.

loan_amount / ltv_ratio / monthly_dti

The calculated loan figures that explain why the loan was approved, reviewed, or stopped.

local_decision / governance_action / final_decision

The app’s decision, the rule check, and the final outcome, saved as three separate facts.

A loan the agent could not approve alone

Morgan’s file from the simulator, followed through the people who touch it.

  1. Credit policy lead

    Sets the rule: any approval with monthly debts above 43 percent of income goes to an underwriter.

  2. Underwriter

    Gets Morgan’s file in the Inbox with income, debts, and the rule that held it, and approves it with conditions.

  3. Mortgage operations owner

    Sees how many files cleared without a person this week, and how long held files waited.

  4. Fair-lending reviewer

    Samples declines months later and finds the reason and rule version on each one.

Where mortgage automation gets stuck

The underwriting agent is rarely the bottleneck. The work around it is.

Conditions and exceptions land in a shared queue with no owner or due date.

Each threshold change waits on a code release, so credit policy and the agent drift apart.

When a fair-lending question arrives, someone rebuilds the file from three systems.

Mortgage work an agent can take on

Start with the decisions your team makes most often and can already describe as rules.

Underwriting decisions

The agent proposes approve, refer, or decline. Lender rules decide whether that stands before the borrower hears anything.

Disclosure timing

Hold a closing step until the disclosure waiting period is met, and keep the dates with the loan file.

Servicing and loss mitigation

An agent proposes a workout option. Exceptions go to a servicing specialist with the payment history attached.

Built-in handoff

Quality control samples a quarter of files

Quality control reviewer

The reviewer works inside an engagement scoped to the underwriting agent and those months, and can see which files are missing evidence.

See the six role portals and how work moves between them

What each lending team gets

Operations

A view of how much of the pipeline moved without a person, and where held files are waiting.

Credit policy

Try a rule change in Shadow first to see which loans a new limit would hold before it applies.

Compliance

Decline reasons and rule versions stay with each file for an adverse-action or HMDA question.

Try it on one loan program

Connect the underwriting agent, load your credit policy as rules, and see which files it clears on its own.

Start free for 60 daysTalk through your workflowTry the portal as each role