Tursio for Credit Unions: Powering Structured Data Search with Automated Context Graph

This paper introduces Tursio, a secure, on-premises platform that empowers credit unions to query complex structured databases using natural language by leveraging Large Language Models to automatically generate semantic knowledge graphs and compliant query plans.

Shivani Tripathi, Ravi Shetye, Shi Qiao, Alekh Jindal

Published Tue, 10 Ma
📖 5 min read🧠 Deep dive

Imagine a Credit Union (a community bank owned by its members) as a massive, ancient library. Inside this library, the books (data) are organized in a way that only the original librarians from the 1970s understand. The books are stacked in a complex, hierarchical maze where finding a single fact about a member requires climbing ladders, deciphering old codes, and asking a specialist to write a custom map for you.

This is the current reality for many Credit Unions. Their data is there, but it's locked behind a wall of complexity.

Enter Tursio: The "Magic Librarian" with a Memory Palace.

Tursio is a new tool designed to break down that wall. Instead of forcing people to learn the library's ancient code, Tursio acts as a super-smart, on-site librarian who instantly understands what you are asking for and finds the answer in plain English.

Here is how it works, broken down into simple concepts:

1. The Problem: The "Gold Mine" Nobody Can Dig

Credit Unions have a "gold mine" of data about their members (loans, savings, transactions). But because their computer systems are old and rigid, getting that gold out is hard.

  • The Old Way: A manager wants to know, "Who are our members with risky loans?" They have to email an IT expert. The IT expert spends hours writing a complex script to dig through the data. By the time the answer comes back, the opportunity to help the member is gone.
  • The Frustration: It's like having a treasure map but needing a PhD in archaeology just to read the first line.

2. The Solution: Tursio's "Automated Context Graph"

Most modern AI tools try to fix this by asking humans to manually teach the computer what everything means. They say, "Hey computer, when we say 'loan,' we mean this specific table." This is like manually labeling every single book in the library before you can search for one. It takes forever and is prone to errors.

Tursio does the labeling automatically.

  • The Analogy: Imagine Tursio is a robot that walks into the library, scans every single book, and instantly builds a giant, 3D mental map (a "Context Graph") of how everything connects.
  • It figures out that "Account" connects to "Member," which connects to "Loan."
  • It learns that "Closed Account" means a specific status in the database.
  • The Best Part: It does this without humans having to write a single instruction manual. It just knows.

3. The Superpower: On-Premises Privacy

Credit Unions deal with sensitive financial data. They can't just send their member data to a public cloud server (like a public library) to be processed.

  • Tursio's Approach: Tursio is like a secure, private vault that sits inside the Credit Union's own building. It never leaves the premises. It processes the data locally, ensuring that member secrets stay secret, while still using the power of advanced AI to find answers.

4. How It Helps Real People (The Scenarios)

The paper shows how Tursio changes the daily life of a Credit Union employee:

  • From "Account-Centric" to "Member-Centric":

    • Old Way: The database thinks in terms of "Account #123."
    • Tursio Way: You can ask, "Show me all members who closed their accounts last quarter." Tursio understands that "Member" is the human, not just the account number, and connects the dots automatically.
  • Speaking Human, Not Computer:

    • You can ask complex questions like, "Find members with erratic spending habits." The computer doesn't have a column called "erratic." Tursio figures out that this means looking at transaction dates and amounts, and it builds the right search query for you. It's like asking a smart friend for advice rather than trying to program a calculator.
  • Fixing Confusion (The "Ambiguity" Problem):

    • Sometimes, a database has two columns both called "Close Date"—one for loans, one for cards. A human might get confused. Tursio spots this confusion, asks you, "Did you mean the loan or the card?" and lets you teach it once so it never gets it wrong again.
  • Trust and Safety:

    • Before showing you the answer, Tursio shows you how it found it (like showing your work in math class). You can see the path it took, save the search for later, and even give feedback if it got something wrong. This builds trust, which is crucial when dealing with money.

The Bottom Line

Tursio is the bridge between the messy, complex world of old banking databases and the simple, natural way humans want to ask questions.

It turns a Credit Union from a place where you have to wait days for a data report into a place where you can ask, "Who needs help?" and get an answer instantly, securely, and accurately. It's not just a search engine; it's a translator that turns raw data into actionable wisdom, allowing Credit Unions to serve their members better than ever before.