F I D U C I A C O R P N O T E · JULY 2 0 2 6

What a Trustee Must Now Be Able to Prove

Artificial intelligence is usually discussed as a technology question. For trustees, that is already the wrong starting point.

The more important question is what happens to fiduciary judgement when machines become capable of researching, comparing, drafting and detecting patterns at a speed no human team can match.

Trusteeship has never depended on technology. It depends on judgement: making decisions where interests conflict, information is incomplete and the consequences may not become apparent for years. AI does not remove that responsibility. It makes the quality of the process behind it more visible.

Much of the current debate focuses on productivity. The gains are real. But AI also has an uncomfortable characteristic: it exposes weaknesses that firms have been able to manage manually for years.

Where information is structured, records are consistent and governance is clear, AI can amplify a strong operating model. Where files are fragmented, documentation varies between offices or important knowledge resides in the heads of a few senior administrators, technology does not remove those weaknesses.

It reveals them.

That leads to a more useful test than asking whether a firm has an AI policy. Can a decision be traced to the information that informed it? Can the underlying sources be verified independently? Can the professional review applied to an output be demonstrated? Can the firm show what was accepted, challenged and rejected? And could someone who was not present reconstruct the reasoning five years later?

If the answer to those questions is uncertain, the AI problem may not be AI at all. It may be governance.

That distinction matters because an AI-generated answer is not evidence that fiduciary judgement was exercised. A model may produce an apparently sophisticated conclusion in seconds. What matters is whether the professional using it can explain why that conclusion was relevant, what was independently checked and why the ultimate decision was reasonable. The evidentiary record cannot stop at the output; it needs to capture the process around it.

This becomes particularly important when AI moves beyond drafting and research into areas closer to discretion: beneficiary assessments, distribution analysis, risk classifications or conflict identification.

Trustees have always relied on advisers and delegates. But even permissive delegation regimes are built around an obvious assumption: responsibility is being allocated to an identifiable person or institution capable of being held to account. An AI system complicates that assumption. It can materially influence a decision without being a legal actor to whom fiduciary responsibility can meaningfully migrate.

The question is therefore not whether AI participated in the process. It is whether the trustee can defend the resulting decision independently of the machine.

AI can assist the decision. It cannot assume responsibility for the judgement.

Confidentiality creates a parallel issue. Fiduciary firms hold information that clients reasonably expect to remain within tightly controlled professional environments. Before that information enters a third-party AI system, firms need to understand where it goes, how it is retained, who can access it and whether its use is compatible with the obligations governing the relationship. That is not merely an IT procurement question; it is part of fiduciary governance.

But there is another side to the argument, and it may ultimately prove more consequential.

Today, most firms are asking whether using AI creates unacceptable risk. The direction of travel suggests that they will eventually have to answer the opposite question as well.

If technology becomes demonstrably better at identifying conflicts, anomalies or inconsistencies that a competent professional is expected to detect, not using it will become increasingly difficult to defend.

That does not mean every new tool should be adopted. It means the risk analysis is becoming two-sided. A firm will need to justify both the decision to use AI and, in the appropriate circumstances, the decision not to use it.

The strongest firms will not necessarily be those deploying the most artificial intelligence. They will be those capable of showing where technology contributes, where its limits begin and where human judgement remains decisive.

Integrity, independence and professional scepticism are not changing. What is changing is the expectation that those qualities can be demonstrated through a clear and defensible record.

Artificial intelligence will not remove the need for fiduciary judgement. It will make it progressively harder to exercise that judgement without being able to prove how it was formed.

Frédéric Sanz

Founder & President Fiduciacorp

Contact

Reach out for tailored fiduciary solutions.

© 2026. All rights reserved.