For the last three years, artificial intelligence (AI) has been at the centre of industry conversations. We’ve seen growing interest in how AI can enhance strategy, transform insurance, and shape the future of advice. More recently, the conversation has evolved from exploring what AI can do to considering what it may be able to do independently. That shift opens up an important new perspective on how we think about the role of AI in our industry.
Growing up in the Free State during the 1980s, AI was already part of daily life on the farm – albeit not this AI. Livestock and stud farmers relied on artificial insemination, and a company called Taurus (now known as Absolute Genetics) led the field.
The parallel with artificial intelligence is oddly instructive: both are technologies designed to improve the odds of success, not guarantee them. Neither is a magic button. Both reward good preparation and punish sloppy input. And in both, someone still has to take responsibility for what results.
Odds, not guarantees
Artificial insemination doesn’t guarantee a calf, but it does improve the chance of conception. Likewise, AI doesn’t guarantee good advice, but it does improve the probability of a better answer.
With artificial insemination, the quality of the biological material matters. With AI, it’s the quality of the data, instructions, and context that matters. For financial advisers, the old rule still holds true: garbage in, garbage out. A brilliant model fed poor client data will produce a beautifully articulated piece of nonsense. Complete, accurate, and product data is therefore not an IT nice-to-have, but rather a condition for AI to be useful at all.
Neither process starts with pressing a button. There is groundwork, selection, testing, and timing. An adviser who wants to use AI effectively needs to understand the client, gather the right information, formulate the right questions, select the appropriate tool, and then review and validate the output. More isn’t necessarily better – piling on prompts, data, and complexity doesn’t automatically guarantee a better answer. Sometimes the fix is better quality, not more quantity.
The right question isn’t how much AI can we use but rather where can AI improve the quality of advice and the client’s experience? That requires selection. You don’t use everything available. You select what has the greatest potential. AI can produce five plausible answers in seconds; the adviser’s job is deciding which one – if any – is good enough. It might be none of the possible answers, and that’s not a failure of the technology. It’s good human judgement.
Artificial insemination assists the natural process; it doesn’t replace the underlying biology. AI assists financial advice; it doesn’t eliminate the need for judgement, empathy, experience, ethics, accountability, or an understanding the client. AI can do the heavy lifting, but it doesn’t sit across the table and understand what keeps the client awake at night.
When AI can act
The conversation is increasingly moving towards agentic AI – systems that plan, use tools, take multiple steps, and, in some cases, act with much less human intervention. A wrong answer on a screen can be corrected. A system with access to email, client information, or other systems can turn a wrong decision into an action. So, the question isn’t only what can AI do, it’s what are we prepared to let it do, and, more importantly, what should we never let it do without human oversight?
Timing matters. Used well, AI is invaluable for preparing for a client meeting, analysing a portfolio, conducting research, explaining a complex concept, or drafting a review letter. Used at the wrong time, it simply accelerates a bad process. The more capable the technology becomes, the more important it is to decide beforehand where the boundaries should be.
The AI conversation has moved beyond productivity and convenience. There is now credible evidence of AI being misused. Anthropic’s own research has documented cases of AI being explored or used in cyber operations and other harmful activities. That should make us uncomfortable, not because AI is inherently bad, but because powerful technology cuts both ways. The same capabilities that can help an adviser analyse information more efficiently can also be used to manipulate it, commit cybercrime, or generate convincing misinformation. AI-generated text, images, audio, and video are becoming harder to distinguish from the real thing.
Both technologies raise questions around consent, privacy, responsibility, unintended consequences, quality control, security, and accountability. If something goes wrong, you can’t blame the technology. The adviser remains responsible for the advice and that principle becomes even more important as AI becomes more autonomous. Delegating a task to AI does not mean delegating accountability. The machine carries no professional responsibility. The adviser does.
The machine has no relationship with the client
AI can analyse a client’s financial position, model retirement scenarios, challenge assumptions, and save hours of work. But it doesn’t genuinely care whether Mr and Mrs Smith can retire comfortably at 63. You, the adviser, do. That relationship is not an obstacle to technological progress – it’s the reason technology should be used to enhance the advice process rather than replace it.
The question was never only about whether the technology is clever enough. It is whether we are responsible enough in how we use it. For advisers, that means learning to work with AI without surrendering professional judgement to it.
AI, like artificial insemination, can occasionally produce an answer you didn’t expect. In both cases, when it works, someone still has to take responsibility for the outcome. AI can make the adviser more capable, but it does not make the adviser less accountable. Ultimately, AI should be an enhancer of human advice, not a replacement for the adviser.
By Hannes Van Den Berg, CEO at Momentum Advice