In financial services, a large share of hours goes into work that consists of fixed steps: checking data, assembling files, building advisory reports, assessing policies or loans against fixed criteria, handling customer questions that have already been asked a hundred times before. That work is usually well documented, because regulators require it. That same documentation also makes it suitable for rewriting into a task that a system can take over or prepare.
What drives the outcome is not the will to innovate but the extent to which the work is already unambiguously recorded, and the extent to which an error is something that can be corrected afterwards or not. A typo in an internal memo is something different from incorrect advice about a mortgage or an insurance policy. Where the consequences of an error are limited and correctable, work shifts faster. Where a decision is binding for the customer or is reviewed by a regulator, a human remains needed for longer to assess the outcome before it proceeds.
Three categories run through almost every process in this sector. Part of the work can be carried out independently by a system: retyping data, sorting documents, answering standard questions based on fixed rules. Part goes with human oversight: the system proposes advice or a calculation, and an employee approves or rejects it, with a documented reason. And part remains human work, because it involves explaining something to a customer in a difficult situation, a judgement that cannot be captured in rules, or a responsibility that must legally rest with a person.
That division is not fixed for the sector as a whole. At one insurer, acceptance of a simple policy has already largely shifted into the second category, with an employee only assessing the exceptions. At a comparable organisation next door, exactly the same task is still done entirely by hand, not because it couldn't be done, but because the underlying data is not in order, the responsibilities have not been documented, or no one has tested whether the system actually recognises the exceptions. The difference is not in the ambition. It lies in what was done before the ambition was voiced.
Management teams in this sector formulate ambitions that often sound clear enough on their own: faster acceptance work, less manual checking, an advisory process largely prepared by the system. The problem is not the ambition, but that four people in the same meeting mean four different things by "faster" or "less manual". What is, for one person, a system that decides independently, is for another a system that only makes a suggestion someone still has to sign off on. That difference determines which capabilities are needed, which decision rights need to be redesigned, and which layer of control must remain intact. As long as that has not been spelled out, people talk about readiness without anyone knowing which readiness is meant.
It is the same pattern that recurs whenever ambition and goal are confused: the difference between an ambition and a goal explains why a statement like "AI-driven acceptance" is a direction and not a measuring point, and why a team must define that difference before it can start talking about hours or capacity.
Readiness is not the question of whether the technology is available. That is usually the case. It is the question of whether the organisation can demonstrate exactly what a task involves, who checks the outcome, which exception must go to a human, and which decision a regulator expects to be traceable to a responsible person. That touches eight dimensions ranging from data quality to decision rights to whether employees are able and willing to fill the new role. An organisation that is far advanced on one dimension — well-documented processes, for instance — may have barely established anything on another, such as clear ownership of exceptions.
Change in this sector also does not move at the same pace everywhere, and that can be seen separately from the ambition itself. Some teams can run a new process within a quarter; others first need a series of decisions and approvals that take months. What differs in this and why is described in how fast an organisation can really change, a question unrelated to how much budget is available.
This framework is not unique to financial services. The same four layers and eight dimensions apply to sectors with a completely different kind of work, as shown by what is happening in construction in terms of AI ambitions via the AI ambitions at play in construction or by what is happening in the AI ambitions in the installation sector. The difference always lies in the kind of work and the kind of error a task allows, not in the approach used to test the ambition.
When work shifts from human work to oversight work or to full takeover, the deployment of capacity changes: hours are freed up, roles shift, some functions take on different content. What an employer does with that — redeployment, restructuring, something else — is a decision for the employer itself and falls under its own statutory requirements around works council involvement and employment law. This page and the accompanying assessment provide no advice on that and no justification for a personnel decision. They provide facts about which work is likely to shift and under what conditions.
The question of which work in your organisation can genuinely be taken over by AI is answered per task with the work scan from FTE TO AI. For those who first want to know where the organisation stands on the eight readiness dimensions, there is the free readiness check: eight short questions, one per dimension, giving a picture of where you are furthest along and where you are least far along. The full ambition assessment, with the four layers and the five confidence gates, is under development.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.