If AI Can Do The Work, What Are Clients Really Paying For?
Nov
I run a wealth management group that advises families across several African countries, many of whom hold assets, run businesses or have children in more than one of them. So when the Financial Times reported on 7 October, citing people familiar with the plans, that HSBC was proposing to cut around 70% of the financial adviser roles in its UK wealth business and about half of its management and specialist roles, I did not read it as a story about one bank. The proposals, which are subject to consultation, form part of a shift towards digital services, AI and operational efficiency. I read them as a statement about what a very large institution now believes clients will pay a person to do.
Angeline Ong, a senior technical analyst at IG, said the cuts showed “AI moving up the value chain into client-facing roles that banks have historically protected.”
One bank’s restructuring is not an industry trend. But when an institution of that size decides it needs far fewer advisers, it has already done the arithmetic on what those advisers were actually doing all day. We should all be asking the same question of ourselves.
Much of what our industry has published about AI this year reaches the same reassuring conclusion: AI won’t replace your adviser, it will make them better. I understand why we say it. I also think it is too comfortable, because it lets us avoid the harder question. AI is not a threat to good advisers. What it threatens is the adviser who was never really giving advice in the first place.
Let’s admit what machines can do
A large part of what our industry calls advice is really information and administration: comparing funds on cost and past performance, rebalancing a portfolio back to its target, explaining how a tax-free savings account works, producing an annual report. Too much of the industry has built its fee model on exactly this work, wrapped it in products and transactions, and called it advice.
Automated platforms have rebalanced portfolios for years, and today’s AI tools can explain how most financial products work, at any hour, in plain language. They will keep improving. Any firm whose value rests mainly on these tasks should be worried, and its clients are entitled to ask what they are paying for.
I welcome that. Every hour an adviser spends on routine work is an hour not spent on the decisions that matter. But if we are going to argue that advisers remain valuable, we had better be able to show why, and not just say so.
What our clients are telling us
In our recent survey of South African clients, 71% said a personal adviser is more valuable to them than ever. In a survey of clients in Zimbabwe, Zambia, Nigeria and Namibia, the figure was 82%.
Clients also want to hear about AI. Some 40% of South African respondents and 57% of respondents in our other surveyed African markets said they would value Carrick’s perspective on it.
The surveys did not ask clients why. My own answer is this: the more information people can get instantly, the more they value someone who can tell them which of it applies to them, and which of it does not.
Their expectations of us will only rise. Having information is one thing. Knowing what to do with it, and carrying the consequences of getting it wrong, is quite another.
AI doesn’t know the whole story
The most widely used AI tools are general-purpose. They can answer questions about almost anything, and some can draw on specialist information and personal data. But they can only work with the information available to them, and they can make mistakes.
That matters particularly for our clients.
A South African investing offshore works within exchange control rules and SARS requirements. A client in Zimbabwe has lived through more than one change of currency in recent years. Consider a family with a business in Lagos, children studying in London, and an investment structure in Mauritius: several sets of rules apply at once, and they interact.
An AI tool may be able to explain each of these issues. It may even identify considerations the client hasn’t thought about. But whether its answer is right depends on the information it has, whether that information is current and whether the rules it applies are the ones that matter. The client may have no easy way of knowing what is missing.
The risk is that AI sounds equally convincing whether it is right or wrong. And even when the information is correct, that does not make the decision right for the person asking. I have seen plenty of financially sound decisions that were wrong for the family making them.
Earlier this year a client showed me the answer an AI tool had given him on moving his money offshore ahead of emigrating. On the allowances it was correct and clearly set out. What it did not know was that he intended to cease South African tax residency, which triggers a deemed disposal of his worldwide assets on the day he leaves, and that he was planning to do it in the same tax year he sold his business. Had he followed the answer as given, the capital gains bill would have run well into seven figures, and he would have found out from SARS, not from the tool.
What’s left for the adviser
According to the reports, HSBC intends to keep human advice for clients with more complex needs. A family whose wealth spans more than one country is exactly that client. If machines take on the information and the administration, what remains is the part that was always the real job:
Knowing what the client hasn’t said. The second marriage, the child who shouldn’t inherit outright, the business partner who wants out.
Holding the whole picture. Looking across borders, currencies and generations, and seeing where the pieces conflict.
Accountability. Carrick gives regulated advice in every jurisdiction in which it operates. That comes with professional obligations, oversight and accountability for the advice we give. None of that can be handed to a general-purpose AI tool. A chatbot does not sit in front of a regulator. That is the difference between providing information and taking responsibility for advice.
Saying no. Talking a client out of a decision at the wrong moment is one of the most valuable things an adviser does, and one of the hardest to automate. Some of the best conversations I have had with clients ended with no transaction at all.
None of this means AI has no place in that work. It can test assumptions, surface information and take administration off an adviser’s desk. But it raises the bar for the adviser. If clients can get the information themselves, we have to ask better questions, challenge their thinking and take a position. Not every advisory firm will manage that. Some were never built to.
What this means for you
Many clients now research financial questions themselves, whether through AI, articles or conversations with friends. We welcome that. Bring what you find to your adviser. A client who arrives better informed is a client we can do more for.
But when a decision affects your family’s wealth across more than one country, currency or generation, ask who understands your whole situation. Who knows what hasn’t been said? Who can see where one decision might create a problem elsewhere? And who is accountable for the answer?
Those questions will matter more with every year. AI will change financial advice. Much of the work we have traditionally done will become faster, cheaper or automated altogether. I am not going to defend the old way of doing things.
I see it as a reason to expect more of ourselves.
At Carrick that is not a slogan. We are building a single platform that takes a new client from first conversation to open account, with the intake, the balance sheet, the proposal and the documentation produced by the system rather than by the adviser at a keyboard. The first phase is in build now and the rest follows. Where we use AI inside it, the tools are constrained by the rules of the regulators we answer to, so that what comes out is compliant before a human ever sees it. The point of all of it is simple: the adviser’s time goes to the client, not to the paperwork. What we will not do is let a machine make a decision for a client. The information can come from anywhere. The advice, and the responsibility for it, comes from a person who knows you.
If technology can do more of the work, then we should have to prove that the advice we give is worth more than the information clients can get for themselves.
The financial advice industry has spent years telling clients why they need us. AI may finally force us to prove it.