Swiss investors: the AI–adviser trust gap
A Swiss survey reports a narrow AI–adviser trust gap among Gen Z investors. The useful response is a clear way to check claims clients bring to their adviser.
A Swiss survey reports only a 0.1-point gap between Gen Z investors’ trust in AI and bank advisers on a five-point scale. The IFZ at Lucerne University of Applied Sciences and Arts published the finding on 28 September. Read the IFZ/HSLU research article.
The study was conducted online in February 2026 with finpension and Demoscope. Its full sample comprised 1,016 people aged 18–79 in German- and French-speaking Switzerland; the researchers describe it as representative by age, sex and language region. The young-investor comparison is a subgroup result, not a description of all respondents. The article sets out the survey method and age-group comparison.
Read the result at the right scale
The published article does not give the subgroup size or uncertainty estimates for the 0.1-point gap. We therefore do not treat it as evidence of statistical equivalence. Nor does this cross-sectional comparison establish that the same people’s trust has changed over time. Those limits concern how the reported result can be interpreted.
Trust also does not establish accuracy. A client’s confidence in an explanation and the explanation’s quality are different questions. Our practical conclusion is to prepare for the conversation in which a client asks an adviser to reconcile an outside claim with the firm’s answer.
Begin with the claim the client wants resolved
Consider a fictional client who arrives with an AI-generated explanation saying that a particular investment can be sold whenever needed. The useful first step is to identify the exact proposition and which product it concerns.
Ask for the relevant excerpt and its date where available, using the firm’s approved channel. Do not ask the client to disclose unrelated personal material from their conversation. Establish the question they need answered: immediate sale, the timing of cash receipt, a withdrawal condition, or something else.
This makes the work specific. “The AI says something different” is too broad to check. “Does this product allow this request under these circumstances?” gives the adviser a proposition they can investigate.
Show the evidence behind the response
We propose a response with four parts:
- The claim checked. State the proposition in plain language and identify any assumption needed to interpret it.
- The evidence used. Point to the relevant current source, document version and passage. Where sources conflict or leave a gap, say so.
- The answer. Explain what the source establishes and what it does not. Distinguish general information from any advice the firm is actually providing.
- The next step. State what, if anything, the client or adviser needs to do before acting. Identify unresolved questions and who will address them.
The same structure should apply if the outside claim turns out to be correct. The purpose is to give the client a supported answer, not to defend one channel against another.
Retain the explanation, not just the appointment
Keep the client’s relevant question connected to the dated response and evidence used, within the firm’s approved record arrangements. If the answer changes, distinguish the new explanation from the earlier one and make clear what prompted it.
Our suggested internal exercise is to use a fictional outside claim with two authorised colleagues. One prepares the response; the other checks whether they can identify the precise question, follow the source and explain the answer without another meeting.
Record unclear terms and missing references. The result is a test of the firm’s response process, not a benchmark of AI accuracy or a conclusion about an individual client’s understanding.
This is where formal communication adds a useful discipline: it makes the firm’s answer attributable, specific and retrievable. The immediate action for a client-service lead is to agree who owns such questions and what a complete response should contain.
For testing whether the explanation itself is understood, see our analysis of the FCA’s disclosure findings.
We reviewed the linked German-language research article. No raw data were reanalysed. This is analysis of published research, not an original Formal Record benchmark. The fictional client scenario and response format are editorial proposals.