USE CASE 01 / BANKING · MORTGAGE REFINANCE
Renewal-at-risk to outbound offer. Under 90 seconds.
The problem.
Rate cycles are unlocking the largest refinance window in a decade.
Tier-1 banks deploy AI propensity models months ahead of renewal to poach books from regional and mid-size institutions.
The mid-size bank doesn't lose on credit decisioning math — it loses on three operational chokepoints.
- —Fragmented borrower signal across core, LOS (Encompass, Blend, nCino), and bureau feeds.
- —Multi-day ingestion of tax docs, paystubs, appraisals, and title.
- —Retention workflows that activate after the competitor's offer has already landed in the borrower's inbox.
The Wedge — BorrowerGraph (banks) / MemberGraph (credit unions).
- —Resolved household-level entity fusing core, LOS, bureau, property, and MLS feeds — lineage-tracked, attribute-level confidence.
- —Event graph that surfaces life events (job change, divorce, inheritance, renewal approach) the moment they appear in any source.
- —Document intelligence pre-trained on mortgage paperwork — IDP throughput from days to minutes.
- —Any propensity, pricing, or retention model the bank runs becomes materially more accurate against the resolved entity than against raw core data.
Engineering wrap.
Agentic orchestration consumes the renewal-at-risk signal, pulls the resolved entity, parses the document bundle, runs explainable pricing, and drops a personalized offer into the bank's digital channel and into the loan officer's queue — under 90 seconds from trigger to outbound.
Credit-union variant — MemberGraph.
The same pattern, tuned to credit-union systems:
- —Symitar
- —Corelation Keystone
- —Jack Henry CUbase
- —MeridianLink
- —Origence
- —dealer feeds
Indirect auto recapture runs on the same substrate.