New Clarity partnered with eMortgage Capital, one of the nation’s largest brokerages, to prototype an AI-powered assistant for loan placement. The prototype ingested a limited set of private lender guidelines and enabled brokers to quickly surface potential fits for complex borrower scenarios, turning hours of manual search into seconds and showing what a scaled solution could achieve.

Overview

For many borrowers, especially those with unique financial profiles, traditional mortgage products do not fit. Brokers often spend hours digging through inboxes, outdated PDFs, and lender portals trying to find a match among private lenders. This slows down deals, frustrates borrowers, and limits the broker’s ability to serve more clients.

New Clarity worked with Zach Spradling, Division Manager at eMortgage Capital, to build a proof-of-concept prototype that automated this process. Using a Retrieval-Augmented Generation (RAG) model, the app ingested guidelines from a select group of providers and surfaced instant answers when brokers provided borrower details.

The result was a working demonstration of how AI could transform loan placement into a faster, more precise, and customer-friendly process.

The Challenge

Brokers at eMortgage Capital faced three core obstacles in serving non-traditional borrowers:

1. Fragmented Information
   Guidelines from private lenders arrive as PDFs and email updates, scattered across inboxes and shared drives. Staying current is difficult even with a small set of partners.
2. Time-Intensive Research
   Finding a home for a complex loan such as self-employed borrowers, unusual income sources, or non-standard credit can take hours or even days of manual searching.
3. Customer Experience Risk
   Every extra day it takes to find a solution increases the risk of losing a customer to another brokerage or leaving them without financing altogether.

The New Clarity Solution

We built a RAG-powered chatbot prototype designed to fit seamlessly into the broker workflow.

1. Guideline Ingestion (Prototype Scope)
   The system ingested guidelines from a limited number of private lenders that eMortgage Capital interacted with most frequently. This provided a test set of real-world data to validate the concept.
2. AI-Powered Search & Match
   Brokers could describe a borrower scenario in plain language (for example, “Self-employed, 2 years in business, 15% down, 640 credit score”). The prototype instantly searched across the ingested guidelines and surfaced relevant programs.
3. Interactive Chat Interface
   Instead of manually parsing PDFs, brokers interacted with a chat app that provided:
   • Matching lenders and loan programs
   • Key requirements highlighted (such as credit score minimums, reserve requirements)
   • Results ranked by fit, based on available data

Implementation Highlights

• Proof-of-concept ingestion of private lender guidelines
• Flowise-powered embedding system to chunk and digest unstructured data
• Pinecone vector database to store embeddings and ensure fast, accurate retrieval
• OpenAI’s GPT API as the “brain” to interpret borrower scenarios and generate natural-language responses
• Flutterflow user interface to give brokers a clean, conversational experience
• RAG (Retrieval-Augmented Generation) pipeline for precise, context-grounded responses

The Results

Even at a limited scale, the prototype showed immediate promise:

• Dramatic Time Savings: Loan placement research that previously took 2–4 hours could be answered in under a minute
• Broker Confidence: Brokers felt more assured they were not overlooking options in their most common lender set
• Customer Experience: Faster responses gave borrowers confidence and trust in their broker

The prototype demonstrated how AI could reshape loan placement by streamlining workflows and unlocking new levels of service quality, pointing toward what is possible with broader adoption.

Testimonial

“What excited me most about this prototype was how quickly it showed the potential of AI in lending. Even though it was built on a limited set of lender guidelines, it gave us a clear view of how powerful the technology could be at scale. Normally, finding the right program for a borrower with non-standard circumstances means hours of digging through emails and lender PDFs. With this system, we could surface those answers almost instantly. That is a game-changer not just for efficiency, but for the quality of service we provide to our clients. It is a glimpse of what the future of our industry could look like.”

— Zach Spradling, Division Manager, eMortgage Capital

Why the Customer Chose New Clarity

• Deep Workflow Understanding – We focused on the exact friction brokers face daily, not abstract AI hype
• Future-Ready Prototype – The project demonstrated what could be possible at scale, with minimal workflow disruption
• Customer-Centric Design – Built to elevate both the broker’s productivity and the borrower’s experience

Possibilities for New Clients

For brokerages, lenders, and financial services firms, AI-driven knowledge systems can:

• Keep teams current on complex, ever-changing product guidelines
• Enable instant answers that free up time for higher-value client work
• Improve customer trust and win rates by delivering faster, more confident guidance

Whether applied to loan placement, underwriting, or client servicing, AI knowledge retrieval unlocks both efficiency and growth.

‍