Ripple has expanded the artificial intelligence tools inside its enterprise treasury platform, positioning the update as a more controlled way for corporate finance teams to use AI without handing over final authority on money movement.
On September 10, 2026, the company said it is broadening GSmart, the AI layer already running inside Ripple Treasury.
The software is designed for the daily work of cash managers rather than as a general chatbot bolted onto finance.
It now covers forecasting, liquidity, risk, reconciliation, and reporting, with each function wrapped in company policy and human review.
The timing reflects a wider industry tension. Large companies are racing to deploy AI agents, but many still lack strong rules for how those agents should operate.
Treasury teams sit at the sharp end of that problem because a bad recommendation can move large sums or breach internal controls.
Ripple’s answer is to keep the arithmetic on deterministic systems while using AI to read policy, spot patterns, and explain what it thinks should happen next.
A person still has to approve every action before anything executes.T
he new release adds three main pieces.
Orchestrated agents watch specific processes, suggest a concrete step such as a cash sweep or a hedge adjustment, quote the exact policy clause that supports the idea, and then wait.
Knowledge Studio lets teams write their own rules and limits so the agents stay inside those guardrails.
Analytics Studio includes Ask GSmart, a conversational tool that pulls answers and insights from the company’s own treasury data.
Early uptake is already visible.
Ripple said 60 percent of eligible customers have turned on Risk Insights, which flags unusual exposures and policy violations.
Forty-four percent are using Forecast Insights, which compares expected cash flows with actual results to highlight emerging shortfalls.
The AI expansion sits on top of Ripple’s earlier move to put digital assets inside the same treasury system.
After buying GTreasury in 2025, the company added native support for holding and managing digital balances alongside traditional cash.
GSmart is meant to add prediction and explanation across that combined picture, so finance teams can see, forecast, move, and put idle balances to work from one place instead of stitching together spreadsheets and separate platforms.
Renaat Ver Eecke, senior vice president of Ripple Treasury, framed the product as “treasury-native AI” rather than generic AI applied to treasury.
The goal, he said, is faster decisions that remain explainable, auditable, and under human control.
For corporate treasurers under pressure to modernize without creating new operational or regulatory risk, the design choices matter.
Calculations stay deterministic.
Recommendations are tied to written policy. Nothing moves until someone signs off. Whether that combination proves useful at scale will depend on how well the agents perform in live environments and how readily teams adopt the new workflow.