# A custom CRM that took mortgage file prep from 8 hours to 25 minutes

Page: https://pixelteh.com/work/alltrust-mortgage-crm · Last updated: 2026-11-19

_Mortgage brokerage · Custom CRM · Canada · 2024 to present_

Alltrust Mortgage had outgrown Zoho, and its brokers spent most of a day on every loan file chasing and checking documents. The CRM we built, with the AI document intake we later added on top of it, took file prep from about 8 hours to about 25 minutes.

## Challenge

Brokers had outgrown Zoho and spent around 8 hours per loan file preparing documents for underwriting, chasing missing ones by email.

## Outcome

A custom CRM with automated document collection, lead scoring and messaging, later extended with AI document intake. File prep fell to about 25 minutes and originations per loan officer rose from 10 to 27 a month.

## Key results

- **8 h → 25 min**: file prep per loan
- **10 → 27**: originations per loan officer a month
- **21 → 13**: days to collect all documents
- **3x**: leads per assistant

## Facts

- **Client**: Alltrust Mortgage
- **Engagement**: Since 2024, ongoing
- **Team**: 2 developers
- **Status**: Live in production, ongoing

## The situation

The brokers worked in a Zoho setup they had outgrown and chased missing documents by email. Two jobs ate most of the time: preparing each file for underwriting, around 8 hours of broker time and sometimes 12, and following up on documents that hadn't arrived.

## What we built

- A data model and workflow engine for leads, borrowers, deal stages, tasks and follow-up sequences, with delays, stop conditions and stage-based automation.
- Automated document collection: secure requests, scheduled reminders, upload tracking and a clear view of what is still missing on every file.
- Rule-based lead scoring that uses financial inputs, current mortgage rates and the brokers' own criteria, plus a suggested next step for each deal.
- Email through SendGrid and SMS and voice through Twilio, with logging and retries.
- Migration of the historical Zoho data into PostgreSQL with full history preserved.

## Results

Collecting a full set of documents now takes 13 days instead of 21, and one assistant handles three times as many leads.

The CRM also became the base for a later AI project: a document intake engine that classifies, reads and validates borrower documents automatically. Together they took file prep from about 8 hours to about 25 minutes, and each loan officer now closes 27 originations a month instead of 10.

## How the engagement evolved

The build started in 2024 with two developers. The CRM has run in production since then with small iterations, and we keep working with Alltrust on the CRM and their partner portal.

## What the client said

> The guys actually knew what I needed. We found common ground pretty fast, and the delivery was smooth. They suggested a couple of different approaches, which we implemented, and proposed extensions. I can recommend them as a partner who actually understands your business and pays attention.
>
> Nick S., CEO, Alltrust Mortgage

## Stack

Node.js, PostgreSQL, Prisma, React, TypeScript, Tailwind CSS, SendGrid, Twilio, Zoho API

## Related

- [Replace spreadsheets with a real system](https://pixelteh.com/services/replace-spreadsheets)
- [Add AI to an existing system](https://pixelteh.com/services/ai-integration)
- [Mortgage and lending](https://pixelteh.com/industries/mortgage-lending)
- Next case study: [Rescuing a live casino, then building its second brand from scratch](https://pixelteh.com/work/crypto-casino-rescue-second-brand): A live casino rescued in under two weeks, then a second brand to 80K users
