The Problem with Monthly Re-Scoring
Portfolio risk teams at consumer lenders in this stage typically know the signal lag exists — Plaid is showing cash flow deterioration weeks before the re-scoring job catches it, and by the time a borrower reaches a loss mitigation queue, early intervention options are narrowing. The $180,000–$350,000 per year in risk ops labor isn't the core problem; the core problem is that the workflow architecture is designed around a monthly batch cadence that made sense before real-time cash flow data was available. The result is preventable delinquencies and charge-offs that an earlier contact could have converted to a modified payment arrangement.
Weekly Re-Scoring Triggered by the Signal Itself
An AI Labor Company agent mines the risk team's existing re-scoring workflow from Snowflake and dbt to understand your deterioration criteria, scoring thresholds, and loss mitigation routing rules. It then deploys an agent that triggers on weekly Plaid cash flow refreshes, re-scores accounts showing deterioration signals against current Experian and TransUnion data, generates proactive hardship outreach via Twilio for accounts crossing your risk threshold, and routes those accounts to loss mitigation — all within the week the signal appears. The Head of Portfolio Risk reviews the routing logic and outreach thresholds; the agent executes the workflow.
Revenue Recovery Is the Primary Mechanism
The business case here is revenue, not efficiency. Earlier contact with at-risk borrowers converts a portion of what would become charge-offs into modified arrangements — that's direct recovery of principal that would otherwise leave the book. Teams in this position find that reducing signal-to-contact lag from 45–60 days to under 7 days expands the window for successful early intervention substantially. The 60–80% reduction in re-scoring lag is the efficiency figure; the revenue case is the charge-off rate improvement that comes from acting during the early signal window rather than after it closes. The agent is typically live in about 4 weeks.
Does the agent send hardship outreach automatically, or does a human review each message first?
The outreach templates and triggering thresholds are set by the risk team. Once configured and approved, the agent executes the outreach workflow — individual message review isn't required, but threshold configuration and template approval are human-controlled.
How does this interact with CFPB and Reg F outreach requirements?
Outreach frequency caps, opt-out handling, and required disclosures are built into the agent's Twilio workflow configuration. Your compliance team reviews and approves the outreach logic before the agent goes live.
We already have a loss mitigation queue in Snowflake — does the agent route directly into it?
Yes. The agent is trained on your existing Snowflake and dbt workflow and routes accounts into your current loss mitigation queue structure rather than creating a parallel system.