Challenge
Evaluating a secondary loan portfolio meant manually building investment-committee-ready company pages across 50-100+ borrowers per deal, pulling from loan tapes, memos, and call notes with no unified data layer — capping the firm's total throughput at roughly a couple hundred applications a year.
Solution
Nimble Gravity built a unified data platform on Microsoft Fabric, paired with an agentic pipeline on Azure AI Foundry that classifies, extracts, and validates deal data before drafting fully-cited company pages with analysts in the loop. The pipeline runs a deployed Claude model, chosen for its strength analyzing and summarizing the portfolio of companies.
Results
3x more deal throughput targeted
from hundred to thousands of portfolio applications reviewed each year without adding headcount
Analyst-reviewed and source-cited
every extracted figure is checked before it reaches the Investment Committee
60-70% less analyst effort per portfolio
freeing time for judgment, not data assembly
All new ingested and historical deal data is searchable
via agentic chat powered by Claude.
