Treasury in the Fast Lane

APIs and AI Take the Wheel

APIs and AI are transforming how treasurers manage money and data. From instant payouts and live balances to AI-assisted onboarding, the pay-off is showing up in faster cycles, cleaner integrations, and better decisions.

For much of the past decade, APIs have been hailed as a breakthrough in corporate-bank connectivity. More recently, AI has joined the hype cycle, promising to reshape everything from cash forecasting to customer service. Yet for corporate treasurers, the reality is more complex, dominated by legacy systems, shifting rules, and the grind of daily operations.

Steven Lenaerts, Head of Global Channels and Digital Onboarding, BNP Paribas, elaborates: “People still talk about APIs as if they’re plug and play, but that isn’t strictly true. There is still an IT integration between two external parties, with its own rules and policies. If it’s executed well, it can feel close to plug and play, but not everyone is at that level of maturity.”

The same applies to AI, which won’t fix weak processes on its own. “AI existed in the 1950s and 1960s,” Lenaerts reflects. “What’s changed is the computing power and the quality of the algorithms. Ultimately, it still comes down to use cases and the problem trying to be solved.”

In both cases, adoption depends on pragmatism rather than fashion. The treasurers gaining real value from APIs and AI are those who start with a clearly defined problem, test the concept in a contained scope, and scale only when the technology adds measurable benefit.

Finding the right fit for connectivity

Despite APIs being part of the treasury conversation for around a decade, they are still far from mainstream in corporate banking, and treasury teams are not yet the largest adopters. “That’s surprising, given the crises we’ve been through when financial supply chains were disturbed,” observes Lenaerts.

Part of the issue is a fragmented treasury technology landscape. Many TMSs are not API-native. Even when corporates implement APIs, the resulting data may feed into a standalone data lake rather than their TMS, which limits the operational benefits.

The wide variation in industry maturity compounds this fragmentation. Digital natives and younger companies with minimal legacy infrastructure are more inclined towards APIs. At the same time, traditional corporates, particularly in complex multi-bank environments, often remain tied to bulk file channels or bank portals for high-volume processes. In practice, many end up with a blend of different connectivity methods.

Hybrid connectivity is emerging as the logical next step for corporate-bank integration: a deliberate mix of APIs, file-based channels, and portals, each used where it adds the most value. “Each connectivity method serves a purpose, and none of them are going away,” Lenaerts reasons. “If you’re sending out 500,000 payments in one go, stick to a bulk channel. Don’t try to use technology just for the sake of it.”

This approach is not a compromise but a way to match the tool to the task, even within a single process. In payment tracking, for example, instead of sifting through a stack of status messages, an API call can pinpoint where a payment is in the chain. “This way of working is far more elegant and straight to the point,” adds Lenaerts.

Shifting gears on payment innovation

While APIs for account reporting were among the earliest use cases, helped by initiatives such as Swift’s account information reporting, it is payments that have taken centre stage in practice.

“If you can digitise a reimbursement or claims process and tie it directly to the payout, you save money and make customers happy,” Lenaerts points out. “It makes the whole process far more efficient.”

Payment-related APIs now span a wide range of applications. They handle reimbursements and refunds in e-commerce and insurance, move funds into and out of digital wallets, and process claims for insurance and warranty services. Increasingly, they are also being used to deliver intraday visibility on accounts, giving treasurers real-time insight into cash positions, and to send credit notifications that automatically reset limits as soon as incoming payments are received.

The operational benefits of linking core business workflows directly to payouts are clear. It eliminates manual steps, reduces error rates and enables treasurers to respond instantly to changes in cash or credit availability.

Pairing APIs with instant payments takes it a step further. “Instant payments with an API is a match made in heaven,” enthuses Lenaerts. In traditional file-based or portal processes, the value of instant settlement can be dulled by the time it takes to trigger and confirm a payment. With an API, corporates can build reachability checks into the workflow, verifying whether a counterparty can receive instant payments and dynamically routing transactions via the fastest, most cost-effective path.

The same logic applies across both domestic and cross-border contexts and can extend to related steps such as fraud screening or compliance checks. The result is an automated, adaptive payments environment that transforms instant rails from merely a speed enhancement into a strategic advantage.

APIs built for the real world

Momentum around payment APIs is transforming the broader API landscape. Early versions concentrated on “plain vanilla” cash management, such as checking balances, initiating payments, and pulling statements. While these functions are useful, they fall short of unlocking the full potential of API connectivity.

The next phase is a shift towards intelligent, segment-specific APIs that reflect the operational realities of different industries. “You go away from the traditional one-size-fits-all APIs to more precise APIs to cater for certain activities,” Lenaerts outlines.

For example, e-commerce marketplaces often require high volumes of refunds, verification of beneficiary bank details, cyber-fraud screening, and assurance that the correct bank account details have been captured. Wrapping all these checks into a single API is more elegant than forcing corporates to call a generic SEPA payment endpoint with a complete ISO 20022 payload.

Industry-specific APIs go beyond basic transaction handling by embedding intelligence in the workflow. Scoring algorithms assess creditworthiness, fraud risk, or payment priority before execution. Pre-validation through IBAN or beneficiary checks prevents failures and the reconciliation issues they trigger. Compliance logic, including AML and sanctions screening, can be integrated into the call, ensuring that payments reach the bank already vetted. “You tailor the API to the process, not the other way round,” says Lenaerts.

Delivering that sophistication requires a deliberate build-and-scale approach. For BNP Paribas, API maturity starts with a stable, standardised offer that works consistently across clients and markets. Once that base is proven, the bank moves into co-creation mode, developing tailored APIs with individual clients and then assessing whether they can be scaled to others.

“All the APIs we’ve built are being used,” Lenaerts notes with a smile. This indicates that APIs have firmly established themselves in the mainstream. In turn, this is reshaping the economics, with pricing models evolving to reflect their status as core infrastructure rather than experimental pilots.

A matter of trust

Like APIs, elements of AI have been around for decades. The current buzz reflects two converging trends: the exponential rise in computing power and the steady improvement of algorithms, particularly in generative AI.

In banking, most AI applications remain internal for the time being. Typical uses include credit analysis and pre-screening of lending decisions, preparing servicing reviews and personalised client meeting material, creating dashboards from multiple systems, and applying ML to detect fraud patterns. Generative AI offers new enticing capabilities that pave the way for use cases with high potential.

What holds back external deployment is a lack of trust. Generative AI can, for example, produce plausible but false outputs, often referred to as ‘hallucinations’. Governance, data quality, and clarity over who can correct a model’s output are critical. It is crucial to keep a human in the loop to ensure proper oversight and control.

“You don’t want to be the bank in the news because your AI agent made a mistake,” warns Lenaerts. Many tools are piloted internally, giving service teams access to AI-powered insights long before any client-facing release.

How treasurers are taking AI for a spin

On the corporate side, AI adoption is gathering pace. Many treasurers are no longer waiting for banks to deliver products. Instead, they are developing in-house tools tailored to their workflows.

Fraud scoring is a common starting point for corporate-built AI tools. In these cases, a treasury team will develop or configure its technology, sometimes in partnership with a vendor, to run AI-powered fraud checks within the company’s systems. This decentralised detection sits close to where transactions are initiated, enabling the AI to flag unusual patterns or behaviours that might be missed by a bank’s central checks, particularly when payments reflect local market nuances.

Stress testing and ‘what-if’ analysis are other growing applications. Some corporates are building their own AI models to establish a baseline cash forecast and simulate the impact of changes in receivables, payables, FX rates, or interest costs. This strengthens treasury because it enables teams to incorporate more variables and test them in multiple scenarios. “AI can give treasurers decision-making tools that can look at more factors, faster, and with greater consistency,” Lenaerts points out.

AI’s ability to combine data from disparate sources is also being used for KYC compliance. Corporates are deploying it to pull information from different systems, verify documents and streamline submissions to banks or regulators, cutting turnaround times, and reducing rework. Document vetting is a particular advantage, such as determining whether an ID copy is genuinely certified, without requiring a manual check.

In FX risk management, treasury teams are utilising AI to enhance exposure forecasting, generate automated hedging suggestions, and conduct scenario analyses to identify the most effective strategies. “This is part of a broader shift towards pairing predictive accuracy with proactive strategy design, giving treasury a stronger hand in risk management decisions,” says Lenaerts.

Tools that earn their place

Whether working with APIs or AI, Lenaerts’ advice to treasurers is consistent: start with the challenge, not the technology. “These technologies are seductive, but to get started on the front foot, it is crucial to identify use cases where they might help, ensure you have the right skills, prove the value, and then scale,” he advises.

Too many projects stall because they attempt to do everything simultaneously. The treasurers who see results focus on fit-for-purpose design, integrating the appropriate connectivity or intelligence with the processes where it will have the greatest impact.

Used in the proper context, APIs and AI work as precision tools; used badly, they disappoint. The difference lies in knowing when, where, and how to use them.

“Innovation is not about doing something different for the sake of it,” Lenaerts concludes. “It’s about solving a problem better than you could before.”