Could artificial intelligence really supplant financial advisers? The spectre of replacement has been haunting many sectors but, as the AI giants turn their attentions to financial services, it is becoming an increasingly insistent concern for advice groups. Still, a closer examination of online ‘advice’ suggests the profession has no need to panic just yet.
The issue for many advisers has been that potential clients can obtain apparently reasonable financial advice using AI chatbot assistants, such as ChatGPT, Claude or Gemini. Indeed, research consultancy the Lang Cat has found around a quarter of adults have used AI for financial help in the last year – and more than half of them did something with the information they received.
Prompts such as ‘How should I invest for the long-term?’ or ‘How should I structure my retirement income?’ do lead to coherent answers that make appropriate nods to areas such as spending needs, diversified income and a proper understanding of associated risks.
The Lang Cat recently ran an experiment – designed by the University of Edinburgh Business School – that set out to test if the ‘advice’ offered by the so-called large language models or ‘LLMs’ was: a) consistently good; b) consistently bad; or c) not consistent at all.
“As things stand, AI has a PR problem – people use it, but they don’t like it – and the financial advice sector reflects this contradiction.
Different prompts, different ‘advice’
Having given everyone at a conference audience of more than 230 advisers access to one of ChatGPT, Claude or Gemini, the Lang Cat analysts released a client scenario over three rounds – starting simply enough before growing steadily more complicated – and let the different tables of advisers do what they wanted with the various LLMs.
The only rules, the authors of the experiment explained afterwards, were that the advisers had to don a “veil of ignorance” – in other words, pretend they knew nothing about financial services and were not “a bunch of industry pros” – and they had to arrive at answers to certain questions or tasks in each round. A panel of planners, meanwhile, were used as a control group.
The Lang Cat identified real inconsistencies between the individual LLMs and also for different prompts, noting: “In the first round, for example, we created a scenario where the answer was very clearly going to involve investing in an ISA. “Claude and Gemini suggested ISA investment 75% of the time while ChatGPT did so 60% of the time. If you are a member of the public opening a chat window at 11pm because you have just come into a bit of money to invest, you cannot rely on what you are being told.”
Interestingly, the authors of the experiment pointed out later, while Claude mentioned seeking out financial advice around 80% of the time, ChatGPT only mentioned the possibility 25% of the time – and Gemini barely mentioned it at all. The experiment also identified a real skew in the AI trio’s investment recommendations – for example, when asked where to invest money, Vanguard came out top “by some margin”, whether that be on a platform or through LifeStrategy. Nevertheless, the LLMs often mentioned platform providers, such as AJ Bell, Fidelity, Hargreaves Lansdown and ii.
Others have found more troubling biases in the way LLMs offer financial guidance. “We have seen many examples of AI telling advised clients that they are paying too much, that they are not fully diversified and that they should only use an advice firm that charges fixed fees,” says Gillian Hepburn, head of adviser solutions at Vanguard. “There was a case that went viral on LinkedIn where a client was paying 1.2% for advice, platform and investment and were told they were paying too much.”
There are also concerns that AI advice is gender-biased – giving women lower-risk portfolios than they do men for the same risk parameters. “The broad theory that women are more risk-averse than men has meant that AI-led recommendations for the same two clients – one male, one female – saw the female put into a lower-risk portfolio,” Hepburn explains.
Business challenges
Then there are the day-to-day business challenges being created by AI. One instance Hepburn highlights is the emergence of AI-generated complaint letters quoting COBS business rules, which are creating real headaches for financial advice firms. Another challenge has been for compliance – for example, voice-cloning technology has caused problems for some advice businesses, with hackers calling account teams and using fake voices in a bid to get funds released.
For the time being, Hepburn says, she is not seeing any evidence that clients are replacing advisers with AI – adding that they “still want to meet face-to-face with a human”. The bigger worry relates to future generations, who are much more comfortable engaging with a chatbot online – and thus much more inclined to take the advice on offer, regardless of its quality.
This summer, the FCA conducted a study of 18 to 40-year-olds who own or are considering owning investments and found that 56% of those canvassed trusted AI tools – significantly ahead of TV and radio (47%), the press (46%) or social media influencers (29%).
Almost two-fifths (38%) said they believed it was fine to make an investment decision based solely on the outputs of AI. More worrying still, approaching half (44%) mistakenly thought AI-generated financial information was regulated while a third believed they would be covered by the Financial Services Compensation Scheme.
Potential upsides
This would suggest there are few good outcomes from the expansion of AI for advice businesses. In the longer run, it may take their clients while, shorter-term, giving them regulatory and compliance problems. That would seem a high price to pay for, say, the ability to transcribe meetings quickly – although there are certainly those who argue some of AI’s potential upsides for advisers have been underestimated.
Research by Dynamic Planner, for example – which last month launched an agentic upgrade to its Digital Advice Platform – has suggested the highest-performing firms were able to increase clients served per adviser by 41%, while maintaining or increasing client ‘financial happiness’ scores.
Many advice firms have made huge strides in technology adoption in recent years but achieving the straight-through processing and connections has not always been easy. Again, AI may provide a solution to this persistent problem, with Dynamic Planner saying it now “imports and normalises” data from more than 150 asset managers as well as 30 platforms, CRMs and back-office systems.
“Connectors to Microsoft 365 and Zoom are now live along with the ability to push all factfind data back to Intelligent Office,” the company adds. “These integrations provide consistent ‘normalised’ data definitions on which firms can base their use of AI and agents.” The technology’s average suitability letter generation time is four minutes while agentic meeting summaries are saving advisers some 70 minutes per meeting.
Productivity promises
Similar claims are made by ‘Claude for Financial Advisers’ from US AI giant Anthropic, which also launched in late October. For the time being, it is a US product based on US regulation but Anthropic asserts it will deliver “a suite of connectors and workflow skills designed to help with research, prep and documentation tasks that occupy an adviser’s time and keep them from client-facing work.”
There may be compliance advantages too. With more detailed transcripts of meetings, advisers can ensure that they are asking clients the right questions, spot patterns in their answers and adjust their processes if necessary. For Consumer Duty-related compliance, this would be a game-changer.
As things stand, AI has a PR problem – people use it, but they don’t like it – and the financial advice sector reflects this contradiction. On the one hand, the technology is providing a real headache for firms, with clients receiving bad advice that needs to be rectified or, indeed, younger generations not seeking out advice at all. On the other hand, it promises a sea-change in productivity for many advice firms. As for so many professions, then, AI brings potential crisis and potential opportunity – but nothing yet is set in stone.
