Traditional credit scores are breaking down. In the US, millions of creditworthy people are locked out of loans simply because FICO scores can’t evaluate real-time financial behavior. @Plaid just took a big step toward fixing that with their Fall Product Release, leveraging foundational models across credit, fraud, and payments. (Important Note: Plaid is the parent company of @twifintech, but that makes this no less exciting 🔥) Here's the breakdown of what Plaid unveiled and why it matters for credit and lending: 1. Smarter Cash-Flow Underwriting Traditional credit looks backward at static snapshot reports. Plaid’s cash-flow tools look at real-time income, spending patterns, and bill payments. LendScore 2: Plaid’s upgraded core credit model delivers 42% stronger predictive performance than traditional credit data alone. Vertical-Specific Models: LendScore Auto: Cut deep-subprime delinquencies by 26% at the same approval rate. LendScore Home Lending: Approved 6.3% more mortgage borrowers without increasing default risk. Instant Link: Speeds up borrower onboarding by enabling consumers to consent and share cash-flow insights across applications in under 2 seconds. 2. Enter "LendScore Arc" (Transformer Models for Credit) Plaid introduced LendScore Arc, its first transformer-based credit risk score. Instead of taking a snapshot of a bank account, Arc treats financial history like a sequence—analyzing the order and timing of transactions (much like how LLMs analyze words). Performance Boost: Delivers a 20% predictive lift on deep subprime and a 24% lift on superprime borrowers over core baselines. 3. Fighting Fraud & Payment Risk via Sequential AI The same foundation models powering credit are now applied to fraud detection and ACH settlement: Plaid Protect: Uses a sequential AI model to analyze the full chain of user interactions rather than isolated events, driving up to a 40% relative improvement in fraud detection over previous baselines. Signal & Guaranteed Payments: Plaid's payment risk tool (Signal) now analyzes sequence patterns to prevent 26% more ACH returns without raising false flags. The future of underwriting isn't a 3-digit static score—it's dynamic, sequential cash-flow intelligence.