By Dmitry Volkov
AI is exposing the limits of banking architecture, forcing a shift from fragmented processes to continuous, system-wide financial execution.
The conversation around artificial intelligence in banking often revolves around how quickly institutions are deploying models and integrating new tools.
I find that framing increasingly misleading.
It is clear that banks are using AI, but the crux of the matter is how they are using it. Using AI to optimize marketing offerings is a very different concept from a fully AI-native bank that optimizes financial decisions for its customers.
Why AI struggles to move beyond isolated use cases in banking
Investment in AI across financial services has accelerated sharply, with capital increasingly directed toward infrastructure that enables real-time decision-making. The expectation is clear. AI will improve efficiency, reduce costs, and enhance responsiveness. And in controlled environments, it already does.
The difficulty, which many miss, emerges at the system level. Most banks operate on architectures built around fragmentation. Data is distributed across multiple platforms, based on sequential workflows and decisions organised around human checkpoints. These systems perform adequately when activity is discrete and coordination is manual, but begin to break down when the need for automated decision-making becomes continuous.
I believe this is why many AI initiatives produce uneven results. Models can generate insights, but they cannot act freely across systems. Execution still depends on coordination layers that were never designed for it. As a result, intelligence is applied locally rather than systemically.
Additionally, banking has historically been organised around products. Accounts, loans, payments, and cards are distributed and monetised individually. Even digital banks, despite improvements in usability, largely preserved this logic. AI-native banks introduce a completely different organising principle.
When intelligence can operate continuously across data and workflows, the system can coordinate financial activity directly. The unit of value shifts from the product (a user acquiring a credit card, for instance) to the outcome (AI choosing the optimal credit card for a particular customer profile), and this requires a new layer within the system.
Increasingly, institutions are developing coordination layers that sit above existing infrastructure and enable interaction across systems in real time. The function of these layers is to connect fragmented processes and allow decisions to move across them without manual intervention. This significantly changes the competitive advantage of a bank, and I don’t perceive many financial institutions as being ready for that transition.
Reinventing the role of the bank
As coordination improves, the role of AI begins to change. In traditional environments, it supports decision-making. In more advanced systems, it participates directly in execution by monitoring conditions, identifying changes, and triggering actions within defined parameters.
The emergence of autonomous agents reinforces this shift, because these systems operate within ongoing processes responding to signals as they arise. The result is continuous execution.
This has broader implications for how banks function. The bank stops being the place where we make decisions. Instead, it becomes the system through which our financial activity is managed and executed. Here, the conversation becomes more complex. Traditional governance models assume that decisions happen at identifiable points, with clear accountability and the possibility of intervention. Continuous execution does not fit neatly into that structure.
Work from the Bank for International Settlements has already pointed in this direction, noting that as AI systems move closer to execution, risk and control can no longer be managed purely through oversight. They have to be designed into the system itself. For many institutions, that is a harder problem than deploying the technology itself.
One of the less discussed implications of this is that for banks, the source of their competitive advantage begins to change. For a long time, banks competed on product design and distribution. More recently, the focus shifted toward data and analytics. AI has reinforced that trend, but it is now pushing it further.
This changes how performance should be evaluated. As AI becomes more embedded in financial workflows and takes ownership for outcomes, banks may be able to monetize only based on real value created. For instance, did a client make a stellar return due to AI-optimized investments? As of now, banks are not really held accountable in many categories. AI is changing that.
A transition driven by economic pressure
Banking remains a high-cost industry, with inefficiencies rooted in fragmentation and the need for ongoing coordination. AI offers a path to reduce those inefficiencies, but only if it can operate across the system rather than within isolated functions.
According to McKinsey & Company, AI has the potential to generate substantial value in financial services, particularly through improvements in decision quality and operational efficiency. Realising that value, however, depends on more than deploying models. It requires systems that allow those models to act. This is beginning to shape investment decisions. As I previously mentioned, capital is increasingly directed toward infrastructure that enables coordination and execution, instead of incremental improvements at the feature level.
As a result, institutions face a structural choice. They can continue to layer AI onto existing systems, accepting incremental gains, or they can redesign those systems to support continuous execution.
In this regard, the transition toward AI-native banking is underway, but it is uneven. Some institutions are already restructuring their systems to enable real-time coordination and execution. Others remain focused on integrating AI into existing architectures. At first, the difference between these approaches is not always visible. Both can point to progress. But over time, the gap widens.
Integrated systems begin to compound advantages. They respond faster, allocate resources more effectively, and reduce the friction associated with coordination. Fragmented systems move in the opposite direction, where each additional layer of complexity increases the cost of making decisions and acting on them. This is the early stages of divergence, where institutions begin to operate under fundamentally different structural conditions.
Final thoughts
In banking, AI is making visible structural limitations that have existed for years but were previously manageable. Institutions that treat AI as an incremental upgrade will continue to see incremental results.
Conversely, those approaching this as a system design question operate differently. They are not asking where AI can be applied, but what must change for it to function as intended.
This transition will unfold unevenly. Over time, however, the effects compound. The difference will not come down to who adopted AI first, but to who built a system capable of using it beyond isolated improvements.
About the Author
Dmitry Volkov is a serial tech entrepreneur and AI investor, founder of EVA AI, Molit.ai, and Social Discovery Group, one of the largest social discovery companies globally. He has worked with over 20 venture capital firms, served on multiple startup boards, and continues to build and advise ventures across AI and technology.
Source link
#AINative #Banking #Structural #Reset #European #Financial #Review
