Everything we announced at Plaid Effects 2026

Zach Perret
CEO & Co-Founder, Plaid

We just wrapped up Plaid Effects, our annual customer conference. It was great spending time with customers and hearing what teams are building. One theme came up in nearly every conversation: AI is rapidly reshaping financial services.

We believe the next generation of financial products will require a financial intelligence layer that can understand financial activity at scale. Today at Effects, we announced a new set of analytics products, and foundation models designed to help power that future: across fraud, credit, payments, and financial management. Here’s what we shipped.

New foundation models

We announced a set of foundation models trained specifically on large-scale financial data from the Plaid Network, and using them as the common intelligence layer across our products.

First, we announced a foundation model that provides a deeper understanding of transactions data across Plaid - not just what happened, but why. With this model, we’re seeing a 13% improvement in categorization accuracy while income classification is up by 48%.

Now we’re working on a new model that understands financial activity as a continuous narrative. Instead of looking at transactions as isolated events, this model looks at the sequence of transactions to determine what financial behavior means in context, and how it's changing.

Collectively, these models are an intelligence layer shared across our products so improvements means your products are getting better without you having to do anything else.

Fraud detection beyond your four walls

We announced Trust Index 3 (Ti3), the latest machine learning model powering Plaid Protect. Ti3 uses a dramatically more sophisticated fraud graph to identify modern fraud patterns: including synthetic identities, account takeovers, and coordinated fraud rings, by tracing relationships across devices, identities, bank accounts, and applications in real time. In testing, Ti3 catches up to 41% more fraud at the same false positive rate.

Lending built on real-time financial behavior

We rebuilt Plaid Income with a proprietary transformer-based LLM that develops a deeper understanding of the context behind transactions, identifying earned income with far greater accuracy, achieving 86% precision without sacrificing recall.

We also added new cash flow and network attributes to LendScore, Plaid’s credit risk score, to improve underwriting performance, analytics depth, and adverse action explainability. LendScore improves predictive performance by 25% compared to traditional credit data alone and can reduce risk by up to 41% while maintaining approval rate.

Lastly, lenders can now subscribe to refreshed borrower data over time using the same endpoint they already use for underwriting and income. This gives them insights into borrower financial health after a loan has been originated so they can reassess risk and proactively manage changes over time.

Smarter decisioning for earned wage access

We also introduced the Plaid Cash Advance Index, a new risk model purpose-built for cash advance and earned wage access providers. These products involve high-frequency, real-time underwriting decisions, but most providers still rely on signals from a single connected account. Cash Advance Index uses network-level intelligence to predict likelihood of repayment within 30 days, helping providers approve more of the right users, extend more appropriate advance amounts, and manage risk more precisely.

In a randomized A/B test with a leading provider, the model reduced delinquency by eight percentage points with no reduction in approval rates.

Instant bank payments without risk

We launched Plaid Guaranteed Payments to help companies approve more transactions while managing risk. With Guaranteed Payments, Plaid evaluates every transaction with our risk model, if approved, we guarantee settlement. If a payment fails, we cover the loss and handle recovery.

Companies using Guaranteed Payments are already seeing approval rates as high as 90% on instant funding flows, with implementations in as little as two weeks.

Developer tools built for the agentic era

The way developers build has changed. A single developer with an AI coding agent can now ship what used to require an entire team, so we rebuilt our developer tools for that reality.

We launched Sandbox Studio, a new environment inside the Plaid Dashboard that lets developers quickly generate test users, run scenarios instantly, and inspect live responses in one place. We also introduced the Plaid CLI and a collection of MCP servers so AI agents can interact directly with our sandbox programmatically.

Later this year, we’ll launch an MCP server for banks and financial institutions as well, enabling data partners to manage Plaid integrations directly from tools like Claude and Cursor with automated validations, debugging, and AI-suggested fixes.

Let's build the future of intelligent finance together

Plaid started by building the connectivity layer powering thousands of financial applications and services. Now, we’re building the intelligence layer on top of that network - helping companies understand financial behavior in real time and build more personalized, secure, and intelligent financial products.

Thanks to every customer, partner, and builder who joined us this week and continues pushing the industry forward. Many of the products we announced today are available now or in early access. Reach out to your Plaid account team to learn more.

There’s a lot more coming. Check out the rest of the talks from Plaid Effects.