By Tamsin Crossland, Principal AI Architect, Icon Solutions.
The relationship between Banks and AI has entered a new phase. No longer confined to innovation teams and proof-of-concept projects, AI is becoming increasingly embedded in day-to-day operations, particularly within payment functions.
Banks are currently investing more in AI, and payment operations are where two of the industry’s biggest challenges are coming together. Many payment processes still rely on manual checks and interventions that can be costly, time-consuming, and resource-intensive, often diverting skilled employees away from higher-value analytical and strategic work.
At the same time, banks must improve efficiency without compromising resilience, compliance, or customer experience. That tension is creating a potential opportunity for AI to move beyond task automation and potentially support the orchestration of operational workflows across the bank.
Where operational pressure meets opportunity
The complex nature of payment environments creates significant operational challenges, involving payment repair, exception handling, reconciliation, and client servicing. Managing these activities can be both time- and resource-intensive, requiring teams to make accurate decisions under tight operational deadlines while maintaining compliance, resilience, and service standards.
This provides the perfect example of untapped AI potential. Here, AI can play a transformative role in payment repair and exception handling, helping institutions identify routing issues, resolve failed transactions and streamline investigations more efficiently.
Many banks are also experimenting with how AI can deal with client query resolution, by automating payment status enquiries and surfacing transaction data more quickly for operations and servicing teams. This is still, fundamentally, task automation. But the groundwork it creates, through cleaner data, trusted outputs and well-understood processes, is enabling a more significant shift in expanding AI’s role within a Bank.
ISO 20022: The data foundation for AI
While AI is helping institutions automate operational processes, the industry-wide shift to ISO 20022 is making AI’s role in payments operations even more powerful.
As financial institutions converge on a common messaging standard, they gain access to more structured payments data, creating potential opportunities for enhanced automation, reporting, analytics and AI-driven decision-making.
To categorise the shift to ISO 20022 as simply a tidier way of sending payments would downplay the value it has really brought. The standard is helping organisations create more consistent and consumable payments data models, making it easier to connect information across payment lifecycles, improve reporting and support operational analysis.
As these foundations mature, they may also enable more advanced AI capabilities in areas such as reconciliation, exception management and operational intelligence. ISO 20022 functions as much more than a messaging upgrade, the messaging standard is a key enabler in creating further opportunities for AI-driven enrichment, reconciliation and operational intelligence.
From efficiency to orchestration
While still at a relatively early stage, another hugely important development is the emergence of agentic AI orchestration across internal banking operations.
Banks are increasingly experimenting with coordinated AI agents capable of managing multi-step operational workflows, interacting with systems, retrieving information, escalating decisions and collaborating with employees across different functions.
Although this could bring significant value to banks, successful adoption requires a careful balance between deterministic automation and agentic reasoning. Core payment orchestration demands predictable, auditable and repeatable execution to ensure regulatory compliance, operational resilience and settlement certainty.
In contrast, agentic AI is well suited to less deterministic activities such as investigations and exceptions, payment repair, fraud investigation, regulatory analysis and customer servicing, where it can reason across multiple sources and support decision-making under human supervision. Rather than replacing deterministic payment processing, agentic AI is increasingly deployed around it, augmenting the operational activities that surround the payment lifecycle.
More than just a productivity layer
This trajectory, from payments automation to agentic orchestration, illustrates where AI can enhance banking operations.
As AI use cases continue to mature, the technology is cementing itself as a core architectural capability which can be embedded directly into enterprise operations. The industry is recognising this, as AI evolves beyond being viewed solely as a productivity tool.
The long-term success of AI will be shaped by foundations rather than ambition alone. Clear strategic alignment, strong data quality, future-ready architecture, governance frameworks and operational redesign will ultimately determine which institutions are best positioned to scale AI effectively across the enterprise.
Over the coming decade, the most intelligent institutions will not be those that experimented the fastest, but those that built for scale from the outset.
