As the treasury landscape evolves, the focus is shifting from simple automation to agentic artificial intelligence and managing tokenised assets. SEB’s experts explain how data quality and new infrastructure are redefining cash management.
Efficient cash management is increasingly a question of having the right connections, both for advisory and for technical integration. While 2025 was the year of instant payments, 2026 is becoming the year when these real-time flows become intelligent.
These smarter flows are not only changing the architecture but also how treasury teams work daily. For many of them, the driver is practical: resources are often scarce, and automation is expected to remove repetitive tasks so people can focus on more meaningful work and deliver better service within their own organisations.
Data infrastructure is the foundation for AI automation
The rise of agentic artificial intelligence (AI) marks a significant shift. Unlike earlier tools that followed predefined rules and fixed workflows, agentic systems can take independent action toward defined goals. But the real enabler is not the algorithm itself – it is the strength of the underlying data and process design.
“Whether the goal is visibility, decision‑making or automation, everything depends on the quality of the underlying data and processes”, says Ingrid Åsenius, Data Scientist at SEB’s Financial Strategy Labs.
Even with seemingly complete data, inconsistent definitions, standards, and structures across systems and partners create fragmentation long before automation or AI enters the picture. “Without shared meaning, it becomes difficult to form a reliable, consolidated view”, she explains.
The challenge is rarely technical alone. It’s about agreeing on definitions, clarifying processes and ensuring goals are shared. And this work is often underestimated in both scope and complexity. Alignment can’t sit with one team but across everyone who shares the same data and process. One of the most practical strengths of generative AI today is precisely this: making information easier to access and understand for each stakeholder, lowering collaboration barriers.
From a banking perspective, the data challenge is shared as well. Harri Rantanen, business developer in SEB’s Cash Management, notes that banks are one of the key data sources – and in multi‑banking setups, treasurers often need similar information from multiple providers to consolidate their financial status.
This is why standardisation efforts matter, and why they are not always sufficient. ISO 20022 helps harmonise payment and reporting messaging. Still, API standardisation remains uneven, leaving banks and corporates frequently to collaborate with system vendors and integrators to make data delivery and consolidation work in practice.
From automation to agentic AI – promise, but also new risks
The enthusiasm around agentic automation is growing, but experts remain cautious against assuming it is already solved. Anastasia Varava, Head of Research at the bank’s innovation studio SEBx, points out that agentic automation at scale “hasn’t really happened” yet. She says the limiting factors are often governance, verification and security rather than raw model capability.
Even if integration and APIs are in place, organisations need to be able to trace what an agent did, confirm that the right tools were executed, and verify the correctness of the agent’s actions. Because language models still drive agents, they can hallucinate – and when an agent executes tasks, errors can accumulate rather than remain confined to a single answer.
Cybersecurity is another dimension. Anastasia Varava notes that language models, and therefore agentic systems, can be vulnerable to issues like prompt injection, which makes security controls and monitoring essential.
Rolling out agentic systems also requires a new kind of internal infrastructure. Organisations need control of where agents are registered, what access rights they have, and how they are governed. That means skills and oversight remain crucial, especially as tools evolve rapidly and what is built today may be obsolete within months.
A practical takeaway is that not every process needs agents. Anastasia Varava argues that if a workflow is deterministic, traditional automation may be more appropriate, while agentic systems make sense when decision‑making under uncertainty is required and when the right trade‑off between flexibility and control has been designed.
“We are moving toward a future where treasury operations are not just automated, but intelligent”, says Varava.
Next-generation payment infrastructure – real-time creates new dependencies
The payment world is moving beyond the traditional bank account model. Tokenisation – the process of representing money or other assets as digital tokens on a blockchain – enables programmable transactions that settle instantly and atomically, 24 hours a day.
“Treasurers are now asking whether existing payment rails are sufficient, or if they need tokenised money for instant settlement of assets”, says Rantanen. “This development requires banks to become deeply involved in the corporations’ financial ecosystems.”
But real time is broader than just instant payments. Åsenius points out that if corporates move toward fully real‑time operations, they also need adjacent capabilities – such as real‑time FX conversion, real‑time fraud detection, and real‑time alerts – which can increase complexity even as the goal is better control and automation.
The emergence of Central Bank Digital Currencies (CBDC) and stablecoins – digital currencies pegged to assets like US dollars or euros – is providing new ways to manage liquidity. This is particularly relevant for global companies dealing with multiple currencies and time zones.
One reason tokenised money is attracting attention is settlement. Rantanen highlights that securities settlement has historically been a T+2 process (settlement after two days) and is moving toward T+1, while market participants ultimately want instant, automated settlement in which the transfer of ownership and the movement of money are tightly linked to reduce risk.
The discussion also connects to geopolitics and sovereignty. Rantanen argues that stablecoins are currently dominated by US-dollar-based instruments, which raises questions about Europe’s competitiveness and monetary sovereignty – themes that also appear in debates about dependence on large non‑European AI vendors and the need to ensure resilience in a changing geopolitical environment.
Banks shift from service provider to strategic partner
As these technologies converge, the role of the bank is shifting from financial service provider to an even more important strategic partner for corporations and institutions. This requires SEB and other banks to work closely with clients to ensure their data is ready for an autonomous future – and that integration and governance choices are made with both opportunity and risk in mind.
“Our clients wish for us to help them work more efficiently and take a bigger and more collaborative role in their everyday challenges”, says Rantanen.
