High-Level

Every ZHL loan carries borrower conditions that must be cleared before funding. A core set revolves around income: employed income, self‑employed income, and related borrower conditions like down payment verification and property tax obligations.

Core Problem: Today, employed income verification is a manual grind. Reviewers gather W‑2s, recent pay stubs, VOEs, tax returns, and letters of employment, then reconcile them against a dense underwriting policy and GSE guides. They must normalize pay (salary, hourly, variable comp), check stability and likelihood of continuance, compare year‑to‑date earnings to prior years, and handle gaps, job changes, and fluctuating income.

In the context of Employed Income, the real work is exception-based reasoning of whether the income is “reasonable” and “sustainable”, picking a qualifying income figure, and writing a defensible narrative that ties back to policy. Different people regularly arrive at different qualifying income for the same file, as long as they can justify it.

Does become reviewers

When AI-workflows handle execution, your underwriting staff shifts from doing the work to reviewing it: managing agent outs, assessing judgment, and advising Loan Officers. The result is better work product, fewer review cycles, and a team focused on the work that matters most.

Solution for Employed Income Borrower Condition

Use AI agents to turn “doers into reviewers” for employed income.

Inputs

• W‑2s, pay stubs, VOEs, tax returns, and employment letters.
• The relevant underwriting policy and investor guides (for example, Fannie Mae’s income and employment standards).

Agent swarm

• Extraction agents read each document type, normalize pay frequencies, identify base vs variable income, and detect gaps or trends.
• Policy agents map the raw income story to underwriting rules and exception patterns (stability, trending, continuance, allowable variable income, edge cases).
• Decision agents propose one or more qualifying income figures, each with a step‑by‑step rationale and policy citations. Competing agents can propose different but defensible answers and reconcile to a final recommendation.

Human in the loop

• Fulfillment or underwriting staff receive a structured package: extracted data, calculations, policy references, and recommended qualifying income.