Automating Loan Pricing Configuration with Agentic AI

Overview

Our client is a U.S. banking software provider. Its consultants set up loan pricing rate sheets for client implementations. It uses agentic AI to speed up that setup while keeping consultants in control of every published configuration.

Business Complexity

Consultants re-keyed every rate sheet by hand from Word, Excel, and PDF files, each with its own terminology. At about 20 implementations a month (2-3 sheets each), setup was slow and errors surfaced as wrong rates at origination.

Action

  • Embedded agents in the existing portal to ingest, summarize, map, and configure each rate sheet.
  • Set up variables, tiers, pricing matrices, and rates across absolute-rate, multivariable, and Prime + spread models.
  • Flagged ambiguities and assumptions automatically.
  • Saved configurations only as drafts, so consultants review flagged assumptions and publish, and agents never publish on their own.
  • Applied source comparison and test-loan validation to keep every setup fast, governed, and accurate.

Tech Stack

  • Agentic AI
  • Document Ingestion (Word, Excel, PDF)
  • Human-in-the-Loop Review
  • Source Comparison and Test-Loan Validation

Impact

  • 70%+ reduction in configuration effort
  • Up to 6X faster rate sheet setup (from up to 3 hrs to ~30 min per sheet)
  • 90-95% configuration accuracy target, with >95% validation