Being a mortgage broker requires mastering regulations to identify the most appropriate package for a particular client. I'm not a mortgage SME, but here was a clear information retrieval improvement opportunity:

• Regulations were dense and difficult to master. Staff "down" time was used to create indices in order to improve turnaround time on core tasks.
• Regulations are public, making efficient retrieval an advantage

An LLM is a great tool to solve this issue. However, the target users are services professionals desiring an uncomplicated interface. Implementation and usage posed technical hurdles for them so far. Additionally there are requirements regarding information privacy.

This is where Limitless Horizons Development came in. Our purpose is to help specialists in non-AI domains leverage the new technology, allowing them to focus on their core business. The goal is to bring in new tools for tasks that are friction on the value human staff bring. In this case, staff needed information as quickly as possible to make decisions and the act of information retrieval itself is friction.

The following technical pieces provided the solution:

• Make.com: Used to create a quick prototype with the client, to verify certain choices
• Live stack: Heroku, Python, Postgres, Open AI, Discord: Discord talked to a headless Django app on Heroku, providing a Postgres vector DB as a tool to Open AI, which received and processed the query.

The result was a new capability within a messaging app the staff already had.

Staff are constantly encouraged in the UI to not include personally identifiable information and the full database is on an owned server. But a production environment this would require a more detailed look and also the business’ documented security requirements.

Interested in how AI can help your business? Send me a note and we can explore it together!