AI in financial planning: what 6,000 planners told the FPSB

The Financial Planning Standards Board surveyed more than 6,000 financial planning professionals across 24 territories on how they use AI, and the Financial Planning Institute of Southern Africa has shared the findings for the South African profession. It is the largest study of its kind, and the numbers are worth a planner's attention because they describe your peers, not vendors.

The headline numbers

• 78% of planners believe AI will improve client service, and 60% believe it enhances the quality of advice.
• Two-thirds of firms either already use AI or plan to adopt it within 12 months. Adoption is highest at small firms and very large firms, with the middle lagging.
• 50% view AI positively against just 8% negatively; the rest are undecided.
• 59% see potential to reduce the cost of serving clients, and 60% believe AI can widen access to advice for underserved populations.

What planners actually use it for

The revealing part is not the sentiment; it is the use cases. The leading applications are client communications (41%), data collection (33%), operational efficiency (33%) and risk profiling (30%). Notice what that list is: admin. Planners are not asking AI to allocate portfolios or pick products. They are using it to remove the paperwork that sits between a client conversation and a compliant file.

That matches what we see in South African practices. The adviser's scarce hour is the client-facing one; the hours that AI can give back are the ones spent writing up records of advice, consolidating provider statements and reconciling fee remittances. That is exactly the shape of our financial planning agents: the adviser advises, the agents do the recording, reporting and reconciling, and everything crosses the adviser's desk for approval before it counts.

The two concerns that dominate

Planners are not naive about the risks. The top concerns in the study are data privacy and cybersecurity (47%) and the accuracy and reliability of AI outputs (42%). Both are legitimate, and both are solvable with engineering rather than optimism:

• Privacy. In South Africa this means POPIA by design: knowing exactly which systems client data touches, contractual terms under which it is not used to train models, and adviser approval before anything leaves the practice. Vague vendor assurances are not a data-processing agreement.
• Accuracy. The answer is measurement, not trust. Before an agent goes live it should be run against historical cases the practice has already judged, and scored. An output the system is unsure about should be flagged as a gap, never filled in with plausible text. We have written about this approach in what makes a workflow agentic.

What this means for a South African practice

The global picture says the profession has decided AI is an admin tool first, an advice tool distantly second, and that the firms moving now are small practices, not just institutions. For a South African practice the practical question is narrower: which single workflow eats the most adviser time per week? For most practices we meet, it is the write-up after the meeting. That is why our first agent for this vertical is the Advice Recorder: meeting notes in on WhatsApp, a compliant draft record of advice back for sign-off.

For what a record of advice must actually contain, see our plain-language FAIS guide. For budgeting, see what an AI agent costs in South Africa.