Ripple Treasury Extends GSmart AI as Agent Governance Gap Widens
Ripple has expanded GSmart, the treasury AI now built into the platform it acquired for $1bn last year, adding policy-governed agents across forecasting, liquidity, risk, reconciliation and reporting. The capabilities are already in production across the enterprise customer base, and the pitch to finance chiefs is deliberately narrow: keep the arithmetic deterministic, keep humans on the approval button, and use AI only to interpret, flag and explain.
That framing is a direct answer to a governance problem the numbers make hard to ignore. The expansion also folds a four-decade-old treasury business more tightly into Ripple's digital asset infrastructure, a detail the announcement underplays.
What is GSmart, and whose product is it?
GSmart is the AI layer of Ripple Treasury, the treasury management system Ripple bought when it acquired GTreasury for $1bn in October 2025. GTreasury, founded in 1986 and previously majority-owned by HgCapital, processes in the region of $13tn in payment volume annually for a client base that spans mid-market firms to Fortune 500 names. GSmart itself predates the takeover: it launched under the GTreasury brand before Ripple's ownership, and the platform has since been rebranded as Ripple Treasury.
That lineage matters for reading the announcement. The release presents GSmart as Ripple's treasury-native AI, but the underlying system, the customer base and the AI product all originated with GTreasury. Renaat Ver Eecke, quoted in the release as SVP of Ripple Treasury, was GTreasury's chief executive from 2019 until the acquisition. The expansion is best understood as an established treasury vendor deepening its AI feature set, now under new ownership, rather than a blockchain company building treasury AI from scratch.
How does GSmart handle the governance problem?
The core design choice is to separate calculation from interpretation. Deterministic engines run the financial maths, while AI is confined to interpreting policy, spotting patterns and explaining recommendations. Ripple describes orchestrated agents across forecasting and planning, liquidity, risk, reconciliation and reporting, with each agent monitoring its process, proposing a specific action, citing the policy clause behind it, and waiting for sign-off before anything executes.
Two supporting components sit around the agents. Knowledge Studio acts as the policy and governance layer, letting teams define the organisational controls that proposed actions are checked against before a person is asked to approve. Analytics Studio, which includes a conversational assistant called Ask GSmart, provides the reporting and query interface over treasury data.
Why does agent governance matter to treasury teams now?
The timing tracks a widening gap between how fast enterprises are deploying AI agents and how well they are governing them. Gartner forecasts that the average global Fortune 500 enterprise will run more than 150,000 AI agents by 2028, up from fewer than 15 in 2025. In the same research, only 13 per cent of organisations believe they have the right governance in place for those agents. Gartner analyst Max Goss has warned that blocking agents outright tends to push staff toward ungoverned shadow AI rather than solving the underlying risk.
For treasury, where an unchecked action can move real cash or breach a policy limit, that gap raises the stakes on where AI is allowed to act autonomously. GSmart's answer, keeping humans on every financial decision while automating the surveillance and analysis around it, is the same answer a growing number of treasury vendors are converging on. Rivals including Kyriba and Finmo have made comparable claims about agentic treasury capabilities, so Ripple is competing in an increasingly crowded field rather than defining a new one.
What does adoption look like so far?
Ripple points to uptake among eligible customers as evidence the capabilities are landing. It reports that 60 per cent of eligible customers have enabled Risk Insights, which surfaces exposure anomalies and policy breaches, and 44 per cent are using Forecast Insights, which compares forecast and actual cash flows to flag emerging liquidity gaps. Those figures are drawn from Ripple's own reporting and cover only customers eligible for the features rather than the full base, so they are best read as a directional signal rather than an independently audited adoption rate.
The expansion builds on the native digital asset capabilities Ripple added to the platform earlier this year, which gave treasury teams a single view of fiat and digital balances. GSmart adds the predictive and analytical layer on top. Ripple, which says it holds more than 85 regulatory licences across its business, is positioning the combined product as a way for corporate treasurers to view, forecast, move and earn on both traditional and digital assets from one system. The licence figure is company-stated and has been cited at varying levels across Ripple's own announcements this year.
Why this matters to FinanceX readers
For CFOs and treasurers, the significant development is not another AI feature launch but the governance architecture underneath it. The market is moving from black-box forecasting models toward agents that can act, and the vendors winning enterprise trust are those that can show an auditor exactly which policy clause triggered which recommendation and who approved it. GSmart's calculation-versus-interpretation split is a concrete expression of that standard.
The strategic read for investors is Ripple's continued push from cross-border payments into the office of the CFO. A $1bn treasury acquisition, native digital asset support and now an expanded AI layer point to a deliberate attempt to make Ripple's rails and stablecoin, RLUSD, part of everyday corporate liquidity management rather than a separate crypto workflow. Whether large treasuries adopt the digital asset side at scale, or simply take the AI-enhanced traditional treasury tooling and leave the blockchain features aside, is the question worth watching.
