An ambition that presupposes AI work does not only shift tasks. It also shifts responsibility: for a decision, for advice, for a signature that is no longer put there by a human. That shift is not equally far along everywhere. In one company, AI already assesses part of the applications today and a human only signs off on the exceptions. In another company, the same process still rests entirely with people, not because it couldn't be done, but because no one has established who is liable if it goes wrong. That difference is exactly what this gate asks about.
The legal gate does not ask whether AI may be deployed. That is almost always allowed, within limits. The question is sharper: for the tasks you want to shift, who is liable, who must give permission, and what evidence must you be able to show if someone later asks how a decision was made. This falls into the three categories that run through the entire ambition test. Where AI takes over the task independently, it must be established who is answerable for the result. Where a human approves or rejects with reason, that reason must be traceable. Where the work remains human work, the question simply does not arise.
Not a good feeling, not legal advice stating that "it will probably be fine". Evidence here is concrete: an overview per task of who may make which decision and on what basis, a documented explanation of how an AI outcome comes about for the tasks where that can legally be required, and a division of liability that is not established only once a complaint has already been filed. If that overview is missing, the gate has not been passed, even if the technology works perfectly. Readiness and functioning are two different questions; this gate concerns the former.
This test deliberately does not touch on personnel decisions. Whether and how freed-up capacity is redistributed is up to the employer, with the legal requirements that come with that. What is tested here is narrower: is the legal set-up of the work itself in order before it is shifted.
Companies that have already passed this gate generally have one thing in common: they answered the liability question per task before shifting the task, not afterward. Companies where this is still open often do have the technology in place and have formulated the ambition, but the translation into who is answerable for what has been left with the department that happens not to own that question. This is not a matter of good or bad intentions. It is a matter of whether the ambition has been broken down into tasks with an owner, or whether it has remained at the level of a vision. The difference between a goal and an ambition explains why that breakdown is usually missing: an ambition sounds concrete but only becomes testable at the level of a task.
If this gate is not passed, there are three routes, and none of them is a dead end. The first is to answer the liability question after all before proceeding: who signs, who explains, who bears the risk. The second is to adjust the ambition itself into a form in which AI does partly contribute but the final decision remains with a human, with a documented reason for every rejection. The third is to leave the task where it is for now, as human work, until the legal set-up has been worked out. None of the three is a loss. An ambition that does not pass this gate now may well pass in a quarter's time, if the work on liability and permission has been done by then.
The outcome, incidentally, is often linked to other gates. An organisation that does not know whether its data is good enough for AI runs into the same traceability question as the legal gate: without traceable data there is also no traceable decision. And an ambition that struggles at the legal gate often also struggles with the question of whether departmental goals are working against each other, because compliance and operations then work with different definitions of "done".
All these gates, including the legal one, presuppose that you know which tasks are at stake. Which work in your company can genuinely be taken over by AI, and to what degree of human oversight, is mapped per task with the work scan from FTE TO AI, and that is the basis on which the legal question can only then be answered concretely.
This page describes the question, not the answer for your situation. For an initial picture of where you are furthest along and where you are least far along, there is a free readiness check: eight short questions, one per dimension, giving a picture of where the liability question has already been answered and where it has not. Also useful is to look at whether suppliers can keep up legally and operationally, because part of the liability often lies outside your own organisation. The full ambition test, with all four layers and eight dimensions, is under construction.
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.