Responsible AI

What to say when a client asks what you think about AI

When a client asks what you think about AI, acknowledge what it does well, explain what it cannot know, and connect that limit to the judgment an advisor provides.

It often comes at the end of a meeting, almost as an aside, while the client is gathering their things: What do you make of all this AI? Is it going to replace you?

An advisor does not need to defend the profession or promote the technology. A credible answer does three things: acknowledges what AI does well, explains what it cannot know, and connects that limit to the judgment the advisor provides.

Two common answers miss that opportunity.

The dismissive answer

One answer is to wave the technology away: It is fancy autocomplete. It makes things up. I would not trust it with anything that matters.

The problem is not merely that the statement is too broad. The client may already have used an AI system to explain a difficult concept, organize a decision, or produce in minutes something that once took an evening. They may have seen genuine value alongside the limitations.

Dismissal does not necessarily reassure them. It may instead suggest that the advisor has not examined a tool that is already influencing how clients gather information and form questions.

The breathless answer

The opposite response is equally weak: We are using AI across the practice. It has completely transformed how we work.

That sounds like many other technology claims clients have heard. It may also invite an uncomfortable follow-up: If the system is doing the work, what am I paying the advisor for?

An advisor who describes AI only as a source of efficiency can accidentally reduce the relationship to the production of documents and answers. Those are visible parts of the work, but they are not the whole of it.

The question beneath the question

Dismissal and enthusiasm appear to be opposites, but both can function as ways to close the conversation.

The client may not be asking for a technology forecast. They may be testing whether the advisor understands what is changing, what is not, and where the advisor still adds value.

Put more directly: What is the advisor for when a system can produce a plausible answer almost immediately?

That is a fair question. It deserves a direct answer.

A more credible response

The answer has three parts, and the order matters.

Acknowledge what the technology does well

AI can be useful for structured work such as summarizing, drafting, reorganizing information, identifying omissions, and explaining a concept in several ways. Within an approved workflow and with appropriate review, those capabilities can save meaningful time.

Conceding that plainly establishes credibility. The goal is not to minimize a useful technology in order to protect the profession.

Name the limit precisely

The durable limitation is not simply that an AI system can be wrong. People can be wrong too.

The more important distinction is that the system answers from the question it received and the information available to it. It cannot reliably account for the full context of a client's life, especially facts it was never given, could not verify, or had no reason to recognize as important.

A financial life contains many of those facts: an unstated family obligation, a concentrated position the client does not think of as part of the plan, a likely career change, a private concern about health, or a tension between two goals that look compatible on paper.

An AI system may calculate that retirement is affordable. A conversation may reveal that the client expects to support a parent, leave a business gradually, or keep a home the analysis assumed would be sold.

The answer may be competent and still be incomplete.

Recent research on AI-generated financial guidance reinforces the importance of the inputs. A 2026 working paper by Taha Choukhmane, Tim de Silva, Weidong Lin, and Matthew Akuzawa found that structured prompts containing relevant financial variables and explicit assumptions produced recommendations that more closely matched the researchers' life-cycle benchmark than participant-written prompts did. The results came from simulations using specific model versions, not observed client outcomes, so they do not establish that any recommendation is suitable for an individual. They illustrate a narrower point: a plausible answer can depend materially on what the system was told and what the question left out.

Connect the limit to the relationship

Producing a competent first-pass analysis has become faster and less expensive over time. AI accelerates that trend.

What remains valuable is judgment under incomplete information: discovering what matters, recognizing what does not fit the standard pattern, testing assumptions, explaining tradeoffs, and remaining accountable as circumstances change.

That does not make technology irrelevant. It clarifies where technology belongs.

A version you can say out loud

An advisor needs language that sounds natural in the room. Something close to this:

I use it within our firm's approved systems. It is very good at structured work, and it saves me real time. What it does not have is the full context of your life, including the things we may not know matter until we talk them through. It makes me faster. It does not replace the judgment or accountability you hired me for. If I ever believe that changes, I will tell you.

The point is not to memorize the wording. It is to answer without defensiveness, exaggeration, or a claim that the technology can never change.

The prerequisite

Do not use that answer if it is not true.

An advisor who has not spent meaningful time evaluating these tools cannot credibly describe where they help or where they stop. That evaluation should happen only through systems, data, and workflows approved by the advisor's firm. Client-related information does not belong in an unapproved tool, even when obvious identifiers have been removed.

The purpose of firsthand evaluation is not to adopt every new capability. It is to form an informed view, understand the firm's boundaries, and be ready to explain both.

The next time a client arrives with a conclusion produced by AI, the conversation moves from what the advisor believes about the technology to how its output should be reviewed. A practical place to begin is with the prompt and the information behind it.

Treat the question as an opening

The question is not necessarily a threat. It is an opening.

A client is asking the advisor to explain what remains professionally valuable in the relationship. Few questions create a better opportunity to answer that plainly.

Have an honest answer ready.

This article provides general professional information, not individualized investment, legal, tax, cybersecurity, or compliance advice. Advisors should follow their firm's policies and use only approved systems and workflows.

Primary sources

ValaisOS is building a more enduring operating foundation for private-wealth professionals.

Request early access