For the past several years, Artificial Intelligence (AI) has dominated the mainstream in almost every industry, and corporate treasury is no different. By its nature, treasury is defined by controls, precision, and accountability. If it can be trusted, the integration of AI promises to enable systems to take action and optimise efficiency across all treasury functions.

But can it be trusted?

To answer this question and explore the use of AI, and specifically agentic AI, in treasury, Trade Treasury Payments’ Trade and Technology Editor, Carter Hoffman, spoke with Mark Johnson, Chief Product Officer at Ripple Treasury.

With nearly two decades of experience in product and enterprise fintech, Johnson explained the evolution of rule-based automation to agentic AI, systems that act like “an agent,” working beyond human capacity to detect risk and take action before issues become monetary problems.

The shift from automation to intelligence in treasury

As in other areas of finance, the most widespread AI integration in the treasury sector is automation. To eliminate repetitive tasks, reduce spreadsheet dependency, and accelerate reconciliation, treasury automation focuses on delivering efficiency and mitigating risk. According to Johnson, this automation model resonates with CFOs and treasurers, as treasury relies on rules-based and logic-based mechanisms.

The evolution of AI is pushing treasury systems to become more agentic. Following the rise of ChatGPT, Claude, and other large language models, it is generally accepted that most consumers expect products to respond to and work with natural language. This shift in expectations pressures treasury to evolve beyond rules-based workflows toward contextual awareness, capable of interpreting data and generating narrative output.

Beyond efficiency, treasury intelligence also improves critical dimensions such as security and governance. Rather than executing predefined rules, intelligence models can monitor liquidity positions and currency exposures in real time. This allows the system to detect breaches, forecast variances, and mitigate risk before damage occurs.

One example that Johnson provided is around cash forecasting. Typically, treasury teams must spend hours generating forecasts and analysing variances across multiple business units. GSmart Forecast Insights uses AI to diagnose the root causes of variances, generate executive-ready natural-language narratives, and recommend forecast adjustments – all in under 20 seconds.

This shifts treasury analysts from manual report preparation to strategic decision-making.

The hidden barrier to AI in corporate treasury

Despite the excitement around intelligent systems, Johnson identifies a clear constraint that impedes adoption. “It really comes down to three key things,” he said. “One of them is data.” An agentic AI runs on an infrastructure where data is the foundation of its operations. In many corporate treasury environments, data is “fragmented” and “coming from multiple ERPs”, sometimes with “different types of formats.” That is why Ripple Treasury invested early in data infrastructure.

The second barrier is trust, which tends to be shaped by both security and regulation. “Customers have to understand where we stand in this process,” Johnson said. The financial market is typically highly regulated and sensitive to consumers when it comes to products that they are not familiar with, especially with agentic AI. There are a pair of questions that management has to ask: Is the team ready for the new technology? Does our current infrastructure allow us to implement it?

However, the openness of this integration is moving in a positive direction. “A year ago, we were not seeing those levels of questions because people weren’t even ready to get started. Now they’re ready to get started, and they started to do their homework,” Johnson said.

But there are some caveats for buyers. “Not all AI products are built the same,” Johnson warns. Some of them are “wrappers”, systems that run workflows outside core systems. This type of service often exposes sensitive financial data. “The last thing you want [exposed to an unknown third-party system] is your financial data,” he added.

For corporate treasurers, data privacy remains one of the most critical responsibilities they must uphold. In short, the hidden barrier is not whether intelligence is capable, but whether the enterprise can adopt and trust its processes.

Speeding up onboarding and time‑to‑value

Even when treasury teams choose the right system, implementation is often a bottleneck. Treasury management systems (TMS) often require months of configuration, especially when data sources are fragmented and multiple ERPs are integrated. One of Johnson’s arguments was that AI should not be limited to optimising treasury operations post-implementation – it should also accelerate the implementation process.

Interestingly enough, AI is being used behind the scenes to help firms better adapt to AI. At Ripple Treasury, the implementation team is using AI tools to reduce the onboarding time for new clients by half. The system automates data validation across multiple ERPs, intelligently maps data fields, and generates business unit balances for reconciliation, getting clients to value faster.

Treasurers do not want to feel they are stepping into a “black box,” which is why this approach helps build trust. For instance, smaller or resource-constrained treasury teams can highly benefit from faster onboarding. This removes a long-standing barrier that has slowed the adoption of new technologies. Ultimately, time-to-value becomes an essential metric that helps treasurers demonstrate the ROI of intelligence.