Clients Have AI Too: An Attendee's CISI Keynote Experience

Heather reported that client use of AI was among the topics discussed during her keynote experience at the CISI Financial Planning Conference. She highlighted the prospect of clients arriving at meetings with far more information, while warning that they may not recognise risks such as hallucinations, bias and sycophancy. Her response was to encourage planners to prepare for “gentle client education”, drawing parallels with GPs and customer support teams encountering AI-generated material. This opens an important question for the profession: how can that education help clients verify information, exercise their own judgment and recognise when human expertise would add value?

---

The risks she names are legitimate; the allocation of competence in her language is what deserves scrutiny. I would challenge the framing without assuming this message proves a deliberate attempt to counter sovereign OS propositions.

Heather acknowledges your central point: clients will use AI and arrive better informed. But she then frames that development primarily as a problem planners will need to manage.

Three phrases carry the weight:

• “They’ll mostly be unaware” makes a broad claim about client competence without evidence in the message.

• “Blindly trusting” places the client in the role of credulous consumer, with the planner implicitly positioned as the discerning authority.

• “Gentle client education” can mean useful support, but here it establishes a one-way relationship: the professional teaches; the client needs correcting.

The comparison with customer support teams receiving lengthy complaints adds another implication: client-generated material becomes something institutions must handle, rather than potentially useful evidence they should examine. A complaint’s length tells us little about its validity.

The hidden distinction is between educating people to exercise judgment and educating them to defer. Both can sound caring. Only the first builds agency.

Her concerns could support your proposition just as readily. If people need help recognising uncertainty, checking sources, challenging assumptions and knowing when to seek expertise, those are capabilities a sovereign OS should help them develop.

Sovereignty should mean control over one’s records, questions and decisions—not treating AI output as authoritative.

The same scrutiny should also apply to planner-operated AI. Putting a professional between the client and the system does not, by itself, demonstrate that its outputs have been verified. The useful question is: what checking process does each approach provide, and can the client understand and challenge it?

There is a commercial incentive worth examining, although this message does not establish Heather’s motive: an adviser-centred model benefits when greater client access to information is interpreted as greater need for professional supervision. Your model asks whether that access can become greater independent capability.

I would treat her response as an opening. She has acknowledged that clients using AI belongs on the agenda. The next question is:

“Does our response to client AI use make people better at judging information—or more dependent on us to judge it for them?”