Advisers needn’t fear handing control to AI

However, the pace of change is creating some concerns. When I talk to advice professionals about AI, the main worry I hear is the perception that they’re handing over control. The fear is that the client experience will lose what makes it distinctive and start feeling generic, and that there will be no clear evidence of what action has been taken or how decisions have been made.

Take the fact find, for instance. It’s one of the first interactions a client has with a new adviser and, as it varies considerably from firm to firm, it’s a way of reinforcing the proposition. It gives the client a feel for the kind of firm they are dealing with. There’s often a concern that using AI for the fact find will strip out everything that makes it unique to that firm. But by selecting the right mode of AI for the task, it’s possible to take advantage of what it can do while keeping your firm’s points of difference.

The risk and judgement required differ between tasks, so a firm might draft reports using copilot, run meeting follow-ups on supervised and reconcile routine data via autonomous. Firms, or advice professionals, shouldn’t feel forced to give AI more autonomy than they’re comfortable with.

Fixed steps or reasoned judgement

We’ve also baked in additional controls that check whether a process needs to be fixed and repeatable, or flexible and adaptive.

A fixed, or deterministic, process happens the same way every time, so suits anything where a regulator might ask “show me exactly what happened and when”. A flexible, or reasoned, process lets the system adapt to each case as it occurs, which is useful for situations where speed and judgement are more important than rigid repetition.

Again, a firm chooses which approach applies to which process. Compliance-critical steps can stay fixed and fully evidenced, while judgement-heavy steps can flex, with the firm deciding which is which.

The future of AI in advice isn’t about handing control from advisers to machines. It’s about making the level of control explicit: what AI is allowed to do, when a human must approve it, where processes must be deterministic, where reasoning can help, and what evidence must be retained.

Done properly, greater automation shouldn’t mean less control. It should mean firms have a clearer, more deliberate way of deciding where human judgement matters most.

This article was first published in Professional Adviser on 11 September. Please find a link to the original piece here.