Plaid Teaches AI to Understand Finance
Two borrowers can have the same income, the same account balance, and even the same overdraft history, yet represent very different credit risks. Traditional models often struggle to tell them apart. Plaid believes foundation models can.
This year, the company introduced a two-layer foundation model architecture for financial data. The transaction model makes sense of individual financial events, while the sequential model looks at their order, cadence, and relationships, similar to how a large language model understands words in context rather than in isolation.
Improving Lending and Fraud Detection
A richer understanding is improving lending, fraud detection, and payment risk across Plaid‘s network of more than 12,000 financial institutions and 9,000 apps. Its transaction foundation model improved income classification accuracy by 48%, while its sequential foundation model reduced credit default risk by 13.6% at the same approval rate.
This foundation can also feed into products such as LendScore, Plaid‘s real-time cash flow-based credit risk score, which reduced lending risk by 41% compared to a traditional benchmark, without sacrificing approvals.
They see financial activity across 12,000 institutions and 9,000 apps.
Evolution of Plaid’s Approach
Plaid has evolved from a data connectivity company to an intelligence platform. Suddu Seshadri, Plaid‘s Head of Data & AI, notes that the next phase will be intelligent finance, where AI reasons on top of this financial data.
Consumer demand for self-driving money is clear, with 86% of adults in the US already using AI to better understand and manage their money. Plaid wants to ensure its customers have access to the richest data available as they build new intelligent finance experiences.
Financial data has characteristics that make it difficult for traditional models to analyze, such as cryptic transaction descriptions and limited labeled data. Foundation models allow Plaid to learn a reusable representation of financial activity once and adapt it across many tasks.
Related: Sorry, You Have Been Blocked
Building Foundation Models
The company’s approach to building its foundation models is fundamentally different. It combines network data with models built specifically for financial activity, rather than asking a general-purpose model to guess from raw bank text.
The transaction model learns the economic meaning of individual events, while the sequential model learns order, timing, and changes in behavior over time. This shared understanding can then improve product-specific systems across payments, fraud, and lending.
For example, Plaid‘s transaction foundation work has improved income classification and loan payment detection. The shared foundation turns underlying concepts into reusable representations, allowing each product to combine them with its own data, labels, thresholds, and controls.
Suddu Seshadri notes that modern models make it possible to extract more value from context, such as sequence and relationships. A transaction means more when you understand the activity that came before it, its timing, and its relationship to other events.
In practice, this means that Plaid‘s foundation models can help financial institutions and apps make more informed decisions about lending, fraud detection, and payments, ultimately leading to better outcomes for consumers. By providing a richer understanding of financial behavior, Plaid is enabling the development of more personalized, useful, and safe financial products, including online payment systems.
The approach has earned Plaid an award at Tearsheet‘s AI Innovation Awards 2026. As the digital financial ecosystem continues to evolve, Plaid‘s foundation models are likely to play an increasingly important role in shaping the future of financial services.
They are shaping the future.
