AI Inbox Assistant for DSCR Loan Officers: Scans, Status Tracking and Draft Follow-Ups

A loan officer who reads an email and forgets to answer it can lose a deal without ever seeing the loss. On one DSCR lending desk, loan status and messages from borrowers and brokers were scattered between the loan origination system and the email from DSCR broker portals, and the officer admitted to forgetting to reply after reading.

The loan officer originates DSCR and non-QM loans and works with a processor. We built Claude artifacts to scan the inbox, track loan status and draft follow-ups. For mortgage loan officers in the same spot, the order of setup matters more than the tools.

The situation

We mapped the full loan-processing workflow across the loan origination system and the DSCR broker channels. Email was the common thread across all of it, but it was split between more than one inbox.

The scattering was the problem. Status lived in one place, broker messages in another and the replies owed lived only in memory.

What was leaking

Two things were leaking at once. The first was follow-up. The officer read messages and then forgot to answer them, and nothing surfaced the unanswered ones.

The second was visibility. The DSCR emails arrived in a personal inbox, and the automation was connected to the team's shared Gmail. Whatever stayed in the personal inbox was invisible to it. The estimate on the call was that the automation would miss 70% of those emails if they stayed there.

What we built

We built the Claude artifacts during the call and set the surrounding rules so they would have something to read.

• Claude artifacts that scrape the inbox daily, track loan status and draft follow-ups to borrowers and brokers.
• Scheduled runs at 7am, 1pm and 7pm, configured after the call, so the scan happens three times a day without anyone starting it.
• Email notifications turned on in the loan system and in the DSCR portals, because email is the common thread the whole automation depends on.
• Auto-forwarding from the officer's personal Gmail to the team Gmail, so the artifacts can see the DSCR messages.
• A daily digest of unanswered emails from an email digest tool, aimed at the forgotten replies.
• A loan tracker in Google Sheets with columns for the DSCR baseline documents, which the artifact checks for each loan.
• A human review step: the drafts are reviewed and sent by a person, never sent automatically.

What changed

No result was reported, so there is no time saved or deal recovered to claim. Part of the setup was still open when the call ended. During the call the artifact reported that it could not access the sheet, so the sheet permissions had to be finalized before tracking would work. The baseline-document columns also had to be validated, because the artifact checks each loan against them.

What changed is the shape of the day. Three scheduled scans replace a single habit of remembering. Every draft waits for a person to review and send it.

Where to start with your own inbox

This section is advice for your own setup. Find out where your loan emails actually land. If some arrive in a personal address, forward them to the inbox your tools read before you build anything. Turn on every notification in the loan system and portals, since an automation can only act on what reaches the inbox.

Keep a person in the sending seat. A draft saves the writing, and the decision to send stays yours.

Frequently asked questions

Why do auto-forwarding and notifications come before the AI assistant?

The assistant can only read what reaches the inbox it is connected to. If DSCR emails sit in a personal inbox or notifications are off, it misses them. Here the call estimated the automation would miss 70% of those emails without forwarding.

Does the AI send emails to borrowers and brokers on its own?

No. On this setup the artifact drafts the follow-ups and a person reviews and sends every one. The drafts save the writing time, while the review step keeps quality and control with a person.

How often does the inbox assistant run?

It was planned to run on a schedule at 7am, 1pm and 7pm. Each run scans the inbox, updates loan status and prepares draft follow-ups, so nothing waits longer than part of a day before someone looks at it.

What does the loan tracker sheet do?

It is a Google Sheets file with a row per loan and columns for the DSCR baseline documents. The artifact checks each loan against those columns, which only works if the columns are complete and the artifact has permission to read and write the sheet.

If replies and loan status are scattered across inboxes and portals, book a 30-minute AI follow-up and reporting audit. We trace where your messages land and show you the first step to automate.