Financial services largely consists of work related to information: compiling files, checking data, assessing risks, drawing up reports, informing clients about the state of affairs. An advisor, accountant or insurance employee spends a large part of the week collecting and organizing data before a judgment or advice comes into play. The outcome of a service therefore depends heavily on three things: how quickly information is available, how thoroughly it has been checked, and how well it is explained to someone who does not master the subject matter themselves.
That makes this sector sensitive to shifts in what becomes automatic and what does not. Where in a factory the throughput time of a machine is what matters, here the throughput time of a file depends on how much human work still lies between intake and advice.
Organizing and checking data is precisely the type of task that lends itself to being taken over: structured input, recognizable patterns, rules that can be formalized. Part of the file build-up, the initial risk assessment and the report format can be done by a system, with an employee assessing the outcome and approving or rejecting it with reasons. The advice itself, the conversation about a borderline case, the explanation to a client who does not understand something: that remains at its core human work, because it runs on judgment and on trust that does not arise automatically.
This does not happen at the same pace everywhere. A firm that had already standardized its file work can hand off that step to a system more easily than a firm where each file is built up in its own way. The difference does not lie in the willingness to innovate, but in how repeatable the underlying work already was before AI came into view.
As long as file build-up and checking are largely manual work, the provider with the most capacity often wins on speed: more people, more files per week. Once that work is largely done by a system, that advantage disappears. Speed then becomes something anyone with a comparable system can deliver, and the difference shifts to what remains: the quality of judgment on borderline cases, the way advice is explained, and whether a client can verify what an outcome is based on.
The latter carries more weight in this sector than before. A client who notices that a risk assessment was generated by a system wants to know who has still looked at it and on what grounds. Providers who have made that oversight visible and traceable have an advantage that does not come from speed, but from explainability. Anyone who only claims to be faster is defending a claim that competitors with the same system can make just as strongly.
The question on which a firm actually wins therefore shifts along with what remains human work and what does not. Comparable shifts are playing out in construction, where planning and calculation are partly becoming automated, in the installation sector, where fault diagnosis and materials planning are shifting, and in the recreation sector, where booking processes and customer contact are changing. The precise ratio between what has been taken over and what remains human work differs per firm and per service, and cannot be derived from a sector average.
This shift says nothing about what an individual firm should do with its staffing. Whether work that is freed up leads to other tasks, other services or something else is a choice that lies with the organization itself and carries its own legal requirements when it comes to personnel decisions. What is described here is a shift in what the comparison between providers hinges on, not a statement about who should keep or shed which work.
The underlying question of which work in a specific company can truly be taken over by AI cannot be answered on the basis of a sector-wide picture; that question is mapped task by task with the work scan from FTE TO AI. For the comparison with competitors, a different question is relevant: how you know what your company truly wins on, and, because dimensions shift as soon as AI takes over new parts of the work, how often a competitive analysis should be repeated so as not to rely on a claim that no longer holds.
A first step is to write down the points on which you believe you win: speed, price, expertise, service, something else. The free dimension check from competitivebenchmark.net places those claims next to the evidence for them, so it becomes visible which claim holds up and which rests on assumption. The full benchmark, with scores for the entire peer group and an evidence matrix per dimension, is under construction.