GreenBox Loans: Project | GetEducated.ai

Mortgage lending · Project 02

A retrieval-grounded agent over thousands of internal mortgage documents, plus an extraction pipeline that cut a three-day review to under twenty minutes.

Read the case study

Domain: Mortgage lending
Build: 4-month engagement
Corpus: Thousands of indexed documents
Extraction: 3 days → 10–20 min

Project impact

The problem, what we built, and what changes for the business.

01The problem
Critical product knowledge and borrower documents were trapped inside slow, specialist-dependent review.
02What we built
A grounded knowledge agent and a separate document-extraction pipeline moved both workflows into one controlled operating layer.
03The result
Underwriting preparation moves sooner, specialist knowledge becomes reusable, and staff capacity returns to judgment-heavy work.

The brief

Mortgage eligibility, pricing, and policy answers lived across thousands of internal product documents. Finding the right one meant knowing it existed first, and the people who knew were the bottleneck.

Separately, reviewing a borrower's bank statements was a three-day manual process. Statements arrived as PDFs across multiple banks and date ranges, and someone had to read every page to classify credits against debits before underwriting could move.

The deliverables

01
MIA, a retrieval-grounded conversational agent indexing the full corpus (internal mortgage product documents plus competitor product knowledge) behind a single agentic interface for internal staff.
02
Two specialised desks share that corpus: a Greenbox Desk for guideline and policy questions, and a Market Desk for market queries. Per-document scoping keeps an answer inside the sources it should be reading.
03
Regenerate, copy, and feedback controls, plus a standing-by status indicator, so a broker can see the agent is grounded rather than guessing, and flag a bad answer in flight.
04
A separate agentic OCR and extraction pipeline ingests multiple bank statement PDFs in one upload, merges them across date ranges, classifies credits versus debits, and exposes filters by bank, account, statement range, and description.
05
Extractions are named and linked to a client and loan number, so the output feeds downstream underwriting instead of dead-ending in a spreadsheet.

Recorded evidence

Only documented project facts appear here.

3 daysBefore

10–20 minAfter

Bank statement review