Private note

Roger / Nik

Enter the access code to read the note.

Nik Shah → Roger Park / September 2026

Roger,

Loan QC will take a lot of training data and tokens.

How are you thinking about solving that bottleneck?

OPTION AWHAT WE ARE PURSUING
Get an anchor client
Automate processing
Automate underwriting
Train loan QC
Sell recurring QCThen scale third-party QC to other clients
OPTION B
Buy a loan QC servicer
Automate its workflows
Upsell its clients
OPTION C

You have a more creative GTM idea.

The upside of independent loan QC

A useful finding on one loan can become a repeatable review service.

01 / ONE LOAN

Find what the first pass missed.

A decision cites an earlier income document. A later version was in the file. The separate QC reviewer flags the mismatch for a BRG expert.

CITEDIncome v1≠IN FILEIncome v2
02 / THE LOAN BOOK

See whether it keeps happening.

With the lender’s approved files, the reviewer checks for the same issue across a portfolio. BRG experts resolve the findings and separate real defects from false alarms.

03 / THE PRACTICE

Sell the review, then keep improving it.

Start with a paid diagnostic. Add ongoing independent QC, fixes to upstream workflows and periodic review as the lender sees value.

Diagnostic→Ongoing QC→Broader work

For lenders: find repeat defects earlier, focus expert time on exceptions and see where the process needs repair.

For BRG: turn expert judgment into a managed service that can extend to other lenders, while 100x builds and tests the underlying systems with BRG’s team.

Illustrative file and proposed service path. Client scope, data rights, independent review design and measured results would be established in a first engagement.

My team is pursuing Option A.

We have been automating loan processing and are starting underwriting. Humans still need to fix the edge cases and exceptions the system misses. QC stays separate from production, with BRG experts judging the findings.

Nik Shah