Kennisbank

Which work does AI take over first in financial services?

A sector made of information

Financial services largely consists of administrative processing of structured information: assessing applications, mutating policies, checking transactions, completing files, drafting reports. Many of those hours are not spent in the conversation with the customer, but in the steps before and after: retyping data between systems, checking documents for completeness, drafting standard letters, testing applications against fixed criteria. This is work with a recognizable structure and a fixed set of rules, and that makes it different from work where every situation has to be assessed anew.

The sector also operates within a strict framework: supervisory rules, retention obligations, accountability requirements. That framework does not determine whether AI can technically carry out a task, but it does determine whether a company can let that execution run without human control. Assessing an application on the basis of fixed criteria is often manageable for AI; the ultimate responsibility for reporting that assessment to a supervisory authority lies with a person.

Where the hours are already shifting

In parts of the sector you can already see what changes when AI takes over tasks. Recognizable examples:

• Document control for applications: checking whether all required documents are present and consistent is a task AI often takes over, with a person assessing exceptions.
• Initial triage of incoming questions or claims: categorizing by type and urgency can largely be automated, while the substantive handling of more complex cases remains human work.
• Reports and periodic overviews: compiling figures from existing systems is a task where AI does most of the work and an employee approves or rejects the outcome.
• Customer contact for standard questions: frequently asked questions about balance, status or terms can partly be automated, but as soon as a customer presents an exceptional situation, human judgment is needed.
• Fraud and risk signaling: recognizing patterns in transactions is something AI supports, while the decision on follow-up steps remains with an employee, partly because of the accountability obligation that comes with it.

In these examples, the same three-way division keeps recurring: tasks AI performs independently, tasks AI prepares while a person approves or rejects with reason, and tasks that remain human work because judgment, explanation or responsibility is required.

Why it differs per company

Whether and how quickly this shift becomes visible depends not only on the task itself, but on the circumstances in which a company operates:

• The state of the underlying systems. Where data is already structured and linked, a task can be taken over faster than where files still sit separately in different systems.
• The degree of standardization in products and processes. An insurer with many custom policies has fewer tasks that can easily be captured in rules than a provider with a limited, uniform range.
• The supervisory requirements resting on a specific process. Some assessments must demonstrably have been made by an authorized person; that determines where a human remains in the chain, regardless of what is technically possible.
• The size of the organization. With a lot of repetition of the same task, automation pays off sooner than with a small team that handles different files every week.

These factors do not only apply here. In hospitality and in the recreation sector too, the degree of standardization determines how much work can be taken over, even though the work there looks very different in substance from a policy assessment or a risk analysis. And in manufacturing, the state of the underlying systems plays a similar role as in the financial sector: where data is already structured, takeover happens faster than where it still sits separately, as can also be seen in the question of whether AI can take over production registration on the floor.

What this is not

This shift says something about work, not about people. Which tasks AI takes over is a different question from what an employer does with its workforce. Decisions affecting the workforce are subject to their own legal requirements and procedures; that question is not answered here, and this is not intended as a basis for it either. This is about a factual breakdown of work: which tasks AI is already taking over today, which come under supervision, and which remain human work, and what that means in hours.

What you can do now

Whether this picture applies to your own organization depends on how your processes are set up, which systems you use and what portion of your work is already standardized. To get an initial indication of that, there is a free quickscan: twelve questions, no account needed, resulting in an indication of what portion of the hours in your profile can be taken over by AI today. The full work scan, which breaks down a company's work to the level of individual tasks and assesses each one, is still under construction. The quickscan already provides an initial direction, without promising more than an indication.