Plaid Pushes AI Deeper Into Credit, Fraud and Payment Risk

Plaid is expanding its use of AI across lending, fraud detection and payments, with a new generation of models designed to interpret financial behaviour over time rather than rely on isolated data points.

Plaid Expands its use of AI 

The open finance company has introduced new cash-flow underwriting tools alongside AI foundation models for fraud and ACH payment risk, arguing that richer transaction histories can improve decisions where traditional credit or transaction-level analysis falls short.

The move reflects a wider shift across financial services towards models that assess sequences, patterns and context rather than static snapshots.

Cash Flow Data Challenges Traditional Credit Scoring

Plaid’s new lending tools are aimed partly at consumers who sit outside conventional credit-scoring frameworks.

Its Instant Link service allows borrowers who have connected financial accounts through Plaid Consumer Reporting Agency to consent to share cash-flow insights for future applications. Plaid says lenders can access those insights in under two seconds.

The company has also introduced LendScore 2, which it claims is 42% more predictive of a borrower’s ability to repay than traditional credit data alone.

Specialised versions are being applied to particular lending markets. Plaid says its auto model reduced delinquency by 26% among deep-subprime applicants at the same approval rate, while its home-lending model approved 6.3% more borrowers at the same risk level.

A more advanced model, LendScore Arc, uses transformer-based technology to learn from the timing and sequence of transactions. Early testing produced a 20% predictive lift for deep-subprime borrowers and 24% for superprime customers over the core model.

Fraud Detection Shifts From Snapshots to Sequences

Plaid is applying the same principle to fraud.

Its new fraud foundation model, trained on hundreds of millions of data points across the Plaid Network, analyses sequences of activity rather than individual events. Internal testing showed up to a 40% relative improvement over previous baselines.

The model now supports Plaid Protect and its Trust Index scoring framework.

The same sequential approach also powers Signal, Plaid’s ACH payment risk model. In testing, Plaid says the model prevented 26% more ACH returns without increasing false positives.

Guaranteed Payments builds on that analysis by offering more flexible outcomes, including delayed settlement and partial guarantees, rather than a simple approve-or-decline decision.

Financial Context Becomes the Competitive Advantage

The common thread is contextual intelligence. Creditworthiness, fraud and payment risk can all be difficult to assess from a single score or transaction. Plaid’s strategy is to use broader behavioural histories to identify relationships and sequences that may be invisible at first glance.

For lenders, PSPs and payment providers, that points towards a more sophisticated risk model.

The competitive advantage in financial AI may increasingly come not from larger generic models, but from models that understand how real financial behaviour develops over time.