Ripple Treasury expands governed AI tools for finance

Ripple Treasury has expanded its GSmart platform with policy-controlled AI tools for forecasting, liquidity, risk, reconciliation, and reporting, as 60% of eligible customers use its Risk Insights feature.

According to a Ripple Treasury announcement, the updated GSmart system uses specialized AI agents to examine treasury data while enforcing controls set by each company.

Rather than allowing AI models to calculate financial results and act on them without review, the platform separates the underlying calculations from the AI layer. Deterministic engines handle the financial calculations, while GSmart reads company policies, identifies patterns in treasury data, and explains its recommendations.

Finance teams retain control over each transaction because the system requires a person to approve any financial action. Ripple Treasury said the setup gives companies access to more automated analysis without handing final authority to an AI agent.

Ripple Treasury keeps humans in control of transactions

GSmart’s agents focus on defined treasury functions, including cash forecasting, liquidity management, financial risk, account reconciliation, and reporting. Each agent examines information related to its assigned function instead of operating as a general-purpose assistant across the company’s financial systems.

When an agent detects activity that may breach an internal rule, it can alert the treasury team and identify the policy behind the recommendation. Users can then review the relevant clause and the supporting data before deciding whether to proceed.

Ripple Treasury said the feature is designed to make AI recommendations easier to trace. Instead of giving users an unexplained warning or proposed action, GSmart can show which internal control produced the result.

Knowledge Studio acts as the governance layer for the agents. Treasury teams use it to define the policies, permissions, and internal controls that guide how GSmart examines financial information.

Within Analytics Studio, Ask GSmart provides a conversational tool for treasury data. Users can submit questions in everyday language and retrieve information without creating a separate report or manually reviewing several data sets.

The company said its system does not give AI final control over payments or other financial transactions. GSmart can find an issue, explain the supporting policy, and recommend an action, but an authorized employee must approve the transaction.

GSmart adoption reaches 60% for risk monitoring

The expanded features are already available across Ripple Treasury’s enterprise customer base rather than being tested through a limited pilot, according to the company.

Ripple reported that 60% of customers eligible for Risk Insights have activated the feature. Risk Insights searches for unusual exposures and possible policy breaches, giving treasury teams a way to examine issues before approving related activity.

Forecast Insights has reached 44% of eligible customers. The feature compares expected cash flow with actual results and identifies differences that may affect a company’s liquidity planning.

By examining those gaps, the platform can bring potential cash shortages or excess balances to the attention of treasury staff. The employees responsible for the company’s finances can then assess the information and decide whether to move funds, adjust forecasts, or leave existing plans unchanged.

Reconciliation tools form another part of the update. Ripple Treasury said GSmart can help users find differences between financial records while its reporting functions organize treasury information for review.

The company has presented the agents as tools for specific work processes, with policy controls applying across the recommendations they produce. Its approach places automated analysis before the approval stage rather than allowing software to authorize and complete transactions on its own.

AI governance concerns drive demand for policy controls

Ripple Treasury introduced the update as companies increase their use of autonomous software across finance and other business functions.

Citing Gartner research, the company said an average Fortune 500 business could operate more than 150,000 AI agents by 2028. Yet only 13% of organizations believe they currently have the governance needed to manage such agents properly, according to the same research cited in the announcement.

Ripple Treasury positioned GSmart’s policy references and approval requirements as controls for finance departments that must explain how decisions are reached. The model gives employees a record of the policy used by an agent before they accept or reject its recommendation.

“Every CFO is under pressure to embrace AI,” Ripple Treasury Senior Vice President Renaat Ver Eecke said.

Ver Eecke added that financial decisions also need to be explainable, governed, and compliant. GSmart’s design addresses those requirements by keeping the AI interpretation separate from the engines that calculate financial results.

Finance departments often work with payment instructions, liquidity forecasts, debt positions, foreign exchange exposure, bank balances, and other sensitive records. Under Ripple Treasury’s stated model, agents can examine such data and flag an issue, but the company’s authorized staff remain responsible for approving the related action.

Ripple builds GSmart on its GTreasury business

Ripple Treasury grew out of GTreasury, the treasury management software company that Ripple acquired in 2025. Following the purchase, Ripple renamed the business and placed its established corporate treasury operations under the Ripple Treasury brand.

The company said the platform handled more than $13 trillion in transaction value during 2025 across over 1,000 customers. The figure covers activity recorded through the former GTreasury business and does not represent XRP settlement volume.

Separating the figures is important because Ripple Treasury serves corporate finance functions that extend beyond blockchain payments. Its platform covers cash management, forecasting, reconciliation, risk oversight, reporting, and other treasury processes, while Ripple operates separate digital-asset and payment services.

The GSmart expansion adds governed AI functions to those existing treasury workflows. Ripple said the product is already active among enterprise customers, as shown by the adoption rates reported for Risk Insights and Forecast Insights.

In the United States, Ripple Treasury plans to demonstrate the expanded system at Sibos 2026, scheduled for Sep. 28 through Oct. 1 in Miami. The event is expected to bring together banks, payment companies, and corporate treasury teams for discussions covering AI, payments, and digital finance.