Customer: Regional U.S. bank with a growing mortgage business
Teams involved: Compliance, Marketing, Loan Officers across multiple states

The Problem

The bank’s mortgage business was growing faster than its compliance team could keep up.

One compliance officer had effectively become the single subject-matter expert for all mortgage advertising. Every flyer, postcard, email, social media image, and builder co-marketing piece flowed through her inbox — thousands of ads per year, all with different federal, state, and internal rules attached.

A few very specific issues were consuming most of her time.

Reg Z trigger terms were everywhere.
Loan officers loved mentioning “30-year fixed,” monthly payments, “no lender fees,” and promotional rate language. Too often, those ads were missing the required APR, loan amount, payment examples, or “taxes and insurance not included” disclosures.

NMLS and state licensing lines were the constant gotcha.
Ads went out missing the LO NMLS ID, the company NMLS, or state-specific license text. She knew examiners love starting with licensing because it’s the easiest finding.

State examiners would request “all Arizona ads for the last two years.”
That meant digging through email chains, Asana tasks, shared drives, and half-organized design folders, then trying to figure out which were final vs. internal drafts.

Loan officers felt like compliance was being picky.
She was spending hours writing messages like “You can’t say ‘best loan officer’,” “You need the Equal Housing logo here,” or “This phrase triggers a disclosure.” She didn’t want to be the bad cop. She wanted the system to say it.

Co-marketing with builders was risky and fully manual.
Joint ads had to be checked for RESPA compliance, equal promotion, contribution disclosures, and wording that could imply a referral arrangement. Nothing in the workflow helped automate that.

What she really wanted was simple:

A place where a loan officer or marketer can upload any ad whether a flyer, email, postcard, social image, even a billboard and the AI tells them: pass or fail, what’s wrong, and how to fix it. Fast. And then she only has to worry about the edge cases.

The Solution

The bank partnered with Luthor to pilot an AI-powered mortgage advertising review workflow.

1. Self-serve uploading for the field

Loan officers and marketing could upload:

• PDFs (flyers, postcards, builder co-marketing pieces)
• Social images
• Email drafts
• Rate sheets
• Co-branded ads with real estate agents or builders

The system instantly reviewed each piece against federal and state mortgage-advertising rules, plus the bank’s internal “hot word” list — phrases that were prohibited or required specific disclaimers.

2. Automatic checks for the issues examiners look for first

For every upload, Luthor automatically checked:

• Whether Reg Z trigger terms appeared without the corresponding required disclosures
• Whether the LO name had an NMLS next to it, whether the company NMLS was present, and whether any state-required lines or disclaimers were missing
• Whether the Equal Housing logo or legend was required and present
• Whether prohibited “hot words” appeared, such as “guaranteed approval,” “best rates,” or anything the bank had flagged as high-risk
• Whether co-marketing ads properly promoted both parties and avoided RESPA-sensitive language

The AI didn’t just say “fail.” It explained why, highlighted the specific area of concern, and recommended what to change.

3. Role-based control so LOs couldn’t override issues

Loan officers could fix issues and re-upload until the ad passed. Only compliance could override a finding. This meant the field got instant, objective feedback, while compliance retained full authority over exceptions and nuanced interpretations.

4. A clean, final-only archive for state exams

Once an ad passed, it was stored in a vault with metadata for:

• Date
• LO
• Channel
• States of distribution

When a regulator asked for “all Arizona ads from the last two years,” compliance exported only the approved, consumer-facing versions — without exposing internal comments or rejected drafts.

The Impact

Within a few months, the bank’s workflow changed dramatically.

Review times dropped by about 70 percent. Most ads were corrected by the field before compliance ever saw them. Compliance was finally able to focus on genuine edge cases instead of repeating the same rules over and over.

NMLS and licensing mistakes nearly disappeared. The system caught incorrect or missing IDs before an ad hit the public, eliminating one of the easiest examiner findings.

Loan officers stopped feeling like compliance was being picky. The feedback came from the tool, not from a person. The field started trusting the system as the first line of defense, and marketing stopped viewing compliance as a blocker.

State exam pulls became simple. Instead of hunting through years of folders, compliance could produce a complete, state-specific set of final ads in minutes.

Most importantly, the compliance lead went from being the bottleneck to being the quality backstop.

As she put it:

“Now I spend my time checking the AI on edge cases, not babysitting every flyer. The field moves faster, and we’re safer.”