Banks know more about their customers than most technology companies. They can see their customers’ income, expenditure, debt, savings and changes in liquidity almost in real time. Nevertheless, it is increasingly not the institution with the most information that gains the upper hand, but the one that can turn that information into an accurate decision more quickly.
This is a significant difference. A dataset is an asset, but it is only through the right architecture, processes and controls that it can be used to reduce the cost of risk, minimise fraud, improve product pricing or speed up customer service. A bank does not become a technology company simply by launching an app or implementing artificial intelligence. It becomes one when technology transforms the economics of its core business.
A lot of data does not mean a complete picture of the customer
A bank may have a long-standing relationship with a customer, yet still be unable to quickly determine that customer’s actual profitability, total debt or current level of risk. Information remains scattered across credit, payment and accounting systems, digital channels and companies within the same group.
This is no longer solely an IT department issue. Data fragmentation prolongs credit decision-making, hinders price adjustments, increases the number of manual reconciliations and slows down the response to a crisis. The European Central Bank has been pointing out for years that banks’ progress in aggregating and reporting risk data remains insufficient. In some institutions, preparing monthly risk reports took 40 or more working days.
For businesses, this means that a historical customer base does not automatically constitute an advantage over fintech firms. A smaller competitor may have less information at its disposal, but can make decisions more quickly thanks to simpler architecture, common data definitions and fewer intermediary systems.
AI is no longer what sets one bank apart from another
According to the European Banking Authority, 92 per cent of banks in the European Union are already using AI, whilst the remainder are conducting pilot schemes or analysing its applications. Among other things, the technology supports fraud detection, customer service, profiling, anti-money laundering and risk assessment.
Such a high level of adoption is changing the nature of competition. Simply possessing AI models no longer provides a competitive advantage, just as using the cloud or offering a mobile app does not. The same underlying models, computing platforms and analytical tools are available to many institutions.
The difference lies in the data used to train the model, how quickly it receives up-to-date information, and whether its output can be integrated into business processes without weeks of coordination. Even the most advanced algorithm will not improve a bank’s performance if it uses inconsistent data or if its decisions subsequently require manual verification across several departments.
The limitation, therefore, is not access to AI, but the organisation’s readiness to utilise it. The BIS identifies quality, privacy and data security, amongst other factors, as the main barriers. The closer the model is to the bank’s core business, the greater the importance of being able to audit it, explain the outcome and establish accountability.
The key advantage lies in the path from information to decision
The most important measure of a bank’s technological maturity is not the number of records collected or models implemented. It is the time that elapses from the moment information becomes available to the moment action is taken.
In lending, this means the ability to quickly identify a customer’s actual risk. Better data can shorten the process, reduce the number of applications automatically rejected and allow the cost of finance to be tailored more precisely. From the bank’s perspective, this simultaneously impacts revenue, operating costs and portfolio quality.
In fraud detection, value is created within seconds. The bank must identify suspicious behaviour before money leaves the account, but at the same time must not block a large number of legitimate transactions. It is therefore not the model itself that provides the advantage, but the combination of real-time transaction data, customer history, device context and automated response.
A similar mechanism applies to sales. A bank that recognises a deterioration in a company’s liquidity at an early stage can offer working capital financing before the customer begins actively seeking such a solution. However, the same set of data may remain worthless if the information reaches the adviser several weeks later or is used to send an ill-suited marketing campaign.
The real asset, therefore, is not the database itself, but the short, controlled and repeatable path from event to decision.
Data architecture becomes part of the business model
From this perspective, IT modernisation ceases to be an infrastructure project. Its value depends on whether it enables a credit rule to be changed more quickly, a product to be launched, a price to be recalculated, or a response to be made to a new type of fraud.
Simply moving systems to the cloud does not solve the problem if different parts of the bank continue to define a customer, income, exposure or product in different ways. Similarly, building a central repository does not automatically yield results when it is unclear who is responsible for the quality of the information and which source is authoritative.
This means shifting the focus from technological migration to designing a decision-making infrastructure. Data owners, shared definitions, traceability of data origins and the ability to reuse the same information across different processes are becoming increasingly important.
Success metrics also need to change. The number of systems migrated to the cloud says little about a bank’s competitive advantage. More telling are the time taken to implement a new rule, the number of manual adjustments, the availability of real-time data, or the proportion of products utilising a shared information layer.
Regulations do not preclude speed
The most valuable data-driven decisions are also among the most heavily regulated. AI systems assessing the creditworthiness of individuals have been classified as high-risk applications under the EU’s AI Act. This entails additional requirements regarding data, documentation, controls and supervision.
These could be viewed purely as a cost, but a bank equipped with a shared layer for managing consents, data provenance and model monitoring can utilise this infrastructure across multiple projects. Rather than building a compliance process from scratch for each application, it reuses the same control mechanisms.
In such a model, compliance does not come at the end of a project and does not serve merely to approve it. It becomes an element of the architecture that allows new solutions to be scaled up without increasing legal and operational risk each time.
Data will increasingly cease to be the exclusive property of banks
The proposed European open finance framework aims to extend, with the customer’s consent, the exchange of financial data beyond information on payment accounts. The aim of FIDA is to enable other financial institutions and service providers to access a wider range of data, thereby increasing competition and the personalisation of products.
For banks, this means a gradual erosion of the advantage derived from exclusive access to a customer’s transaction history. If information can be transferred to a competitor, value shifts towards its interpretation, the quality of models, ease of use and trust.
This does not mean that data will cease to be important. However, it will no longer suffice as a barrier to entry. A bank that fails to utilise it may end up merely supplying raw data to a more efficient competitor.
A bank cannot hand over the logic of its own business to suppliers
The modernisation of banking is becoming increasingly dependent on cloud providers, AI models, analytics platforms and cybersecurity systems. Whilst this accelerates implementation, it also increases reliance on external infrastructure.
An ECB analysis has shown that around 82 per cent of critical functions outsourced to suppliers are difficult or impossible to replace, and 95 per cent of these would be difficult to bring back in-house.
This is a practical lesson for both banks and technology firms. Providers are no longer assessed solely on the basis of their product capabilities. Data portability, auditability, interoperability, an exit strategy and business continuity are becoming increasingly important. The ECB expects exit strategies for cloud services supporting critical functions to be in place even before they are launched.
A bank may purchase infrastructure and components, but control over data, architecture and decision-making logic remains part of its core competence. Entrusting this to a single provider may accelerate modernisation in the short term, but in the long term it may limit the ability to negotiate prices, change technologies and develop products.
Technology only matters if it changes the bank’s bottom line
For technology firms, the greatest opportunity no longer lies in selling yet another general-purpose AI platform. Banks need solutions that remove specific bottlenecks between data and decision-making: integrating legacy systems, controlling data quality, analysing events in real time, monitoring models and ensuring secure data exchange.
For CIOs, this means a shift from managing infrastructure to co-designing business processes. Data architecture becomes directly linked to product pricing, risk levels, fraud losses and service speed.
For boards of directors, the most important question is no longer how much the bank is investing in AI or the cloud. What matters far more is which decisions, thanks to these investments, are becoming faster, more accurate and cheaper.
A bank is not just a technology company. It still manages its balance sheet, capital, risk and trust. However, an ever-increasing proportion of its performance depends on its technological information processing system. That is why data is not yet a competitive advantage. It only becomes one when it leads to better decisions more quickly and securely.
