Kennisbank

AI maturity at banks, insurers and financial service providers

A sector already accustomed to rules, but not to flexibility

Financial services distinguish themselves from other sectors through a long history of structured data and strict control. Where many organizations still have to start with organizing information, this sector often already has decades of experience with data models, audit trails and risk management. That gives an illusory head start. The data management dimension often scores higher in practice than in other sectors, but that does not mean the organization as a whole is ready for AI. Where, for example, construction struggles with fragmented data across projects, the financial sector tends to struggle with data silos that are so well secured they no longer reach each other.

The relationship between the fundamental dimensions is therefore skewed, but in a different way than elsewhere. IT infrastructure is generally robust, but also sluggish: large, old systems that do not move quickly. Organizational structure is often hierarchical and divided across strict lines of responsibility, which can slow down decision-making about new technology. The result is a sector that scores well on paper on the basics, but in practice is slow with the step from foundation to activation.

Why compliance both helps and holds back

Compliance and risk management are baked into the DNA of the sector. That works both ways. On the one hand, it ensures discipline in data quality and clear delineation of responsibilities, something other sectors still have to build up. On the other hand, the habit of calculating risks first leads to reluctance in deploying new technology before the impact has been fully calculated. That reluctance is not necessarily a disadvantage, but it does explain why pilots more often remain stuck in a test setup than grow into a structural application.

The decision-making dimension is decisive here. In many financial organizations there is no clear overview of who is allowed to make which decision about an AI application, and based on what information. That overview, what belongs in a decision inventory, is exactly where many maturity assessments expose a gap: the technology is present, the authority to work with it is not.

The spread between departments is often greater than in other sectors

What further distinguishes this sector is the size of the difference between departments. A risk department has been working with models and statistics for years, and sees AI as a logical extension of what is already happening. A back office or customer contact department sees AI mainly as something imposed from above, without its own work processes being set up for it. When multiple people from the same organization separately score the seven dimensions, that spread often stands out immediately. The outcome is rarely a uniform picture, and that is precisely valuable: it shows where the conversation about readiness has not yet taken place.

This spread is less coincidental than it seems. Financial institutions are often built up of divisions that historically grew apart from each other, each with its own pace of digitalization. That is a different pattern than, for example, the installation sector, where the spread more often arises between head office and execution in the field, or the real estate sector, where the difference is often between management and development.

Why the order of dimensions cannot be skipped here

The temptation to start with the visible, dependent dimensions is great in the financial sector, because there is already a lot of data and a lot of model experience. Yet the order still applies: organizational structure, IT infrastructure and data management form the basis on which the rest rests. An organization that scores high on paper on data management, but where decision-making and responsibility have not been worked out, runs into the same wall as an organization that still has to start. Why the fundamental dimensions must come first is therefore not a matter of preference, but of the way AI applications actually work: without a solid foundation, everything remains at the level of a pilot, no matter how advanced the model is.

The comparison with other capital-intensive sectors, such as the energy sector, shows that a strong technical foundation does not automatically lead to maturity on the other dimensions. Technology and discipline are necessary, but not sufficient.

From readiness to concrete tasks

The maturity assessment shows whether the organization as a whole has the foundation in place to carry AI responsibly: the structure, the infrastructure, the data management and the decision-making surrounding it. That is a different question than which part of the work itself is eligible for takeover. Once that foundation is mapped out, it becomes relevant to look per task at what AI can take over and what not. That is the domain of the work scan of FTE TO AI, which calculates per task which part of the work can be taken over by AI, based on the nature of the task itself rather than on the readiness of the organization.