Advance of the AI adviser: machine learning’s role in managing fortunes

Artificial intelligence can beat the finest human minds at chess and even at Go, the Japanese board game, where there are more potential moves than there are atoms in the known universe. So setting an AI programme to beat the stock market might seem like a potentially easy task by comparison.

However, while many fund management companies are planning to use AI, they do not think it is ready to take over the entire investment process. “There is no ‘piña colada’ system in which a fund manager can let the AI model loose and then go sit on the beach, sipping cocktails,” says Chris Ford of London-based Sanlam Investment Management.

AI will increasingly be used in specific areas of wealth management, from conducting corporate analysis, through customer research and marketing to handling trades. Over time, these techniques may have a significant effect on the fund management business. However, AI will not be taking over the business and, in the short-term at least, clients may not even notice a huge difference in service.

There is nothing new about using computers for investment. Index-tracking, where funds mimic a benchmark such as the FTSE 100, is a largely automated process. And many fund managers use computer programmes to try to beat the market, usually by analysing previous trading data to spot cheap securities or trading patterns that might be repeated.

In the 1980s, Jim Simons, an award-winning mathematician, set up Renaissance Technologies, a hedge fund which used sophisticated statistical analysis to earn an annual return of 37 per cent in its Medallion fund after its (very substantial) fees between 1988 and 2021. At every moment, computers around the planet are combing the data looking for profitable opportunities, not just in equities but in bonds, commodities and every other conceivable asset.

However, as they conduct their research, these computer programmes alter the very market conditions they are studying. If the computers decide, say, that individual stocks that have dropped by 50 per cent in the past 12 months are now cheap, they will trigger a surge of buying orders. The price will rise and the cheapness will evaporate. Those who bought first will do well, but not those who were last in the queue. In turn, that could trigger a wave of “sell” orders that will drive the price back down.

That brings us back to a problem that dogs fund management, whether it is conducted by humans or machines. The index performance represents the return of the average investor. So by definition, not every fund manager can beat the market. In a world where most fund managers were using AI, then these programmes would be matching wits against each other. Like a game of chess between infallible players, the result could be stalemate. Even Renaissance produced negative returns on its public funds in 2020 (Medallion is no longer open to the public) and suffered investor redemptions.

Another reason why it may be a while before AI takes over the market is that the technology is currently unreliable. There have been some well-known examples of large language models (such as ChatGPT) producing false information; a US lawyer used the programme to compile a court brief, only to find that the precedents it cited were completely made up.

Betterment is a US investment firm that offers “robo advice”, largely to retail investors. John Mileham, the group’s chief technology officer, has worked in the tech sector for more than 25 years, including a role in co-founding a savings app. He says that some of the large language models have a “hallucination rate” and that until that declines, using AI to give investment advice “carries risks when you are a fiduciary”.

Some rich clients may fear that AI will result in another change; the disappearance of human relationship managers, often seen as the heart of the sector. That is not likely to happen in the short term. But traditional fund management companies desperately need to improve their technology, which lags behind the kind of service offered by the new breed of fintech companies.

“The wealth management value chain is rife with cumbersome, time-consuming, error-prone manual processes,” says a report by consultants Capgemini. The report published this summer found that only one in two rich clients are satisfied with their digital interaction with fund managers.

Fund management companies hope that AI can allow many client queries to be answered by “chatbots”. Andreessen Horowitz, the Silicon Valley venture capital group, says a large language model that has been trained on past conversations with customers, along with product specifications, should be able to instantly answer all questions about a company’s funds.

At Bernstein Wealth Management, Tuppence Russo, head of client service infrastructure, says the company developed a chatbot for its advisers and sales team. The chatbot took around six to eight months to develop, as the firm needed to ensure the advice it offered was accurate. Bernstein offers some products based on alternative asset classes, where the details can be very complex. When clients call in with a tricky question, the advisers can now consult the chatbot for an immediate response, asking it, for example, “What kinds of assets does the fund invest in?”

The use of chatbots to interact with clients directly — already adopted by Vanguard, for example — will become more prevalent, which is unlikely to bother younger investors. “About 70 per cent of people in Gen-Z are fully digitised and [so are] 60 per cent of millennials,” says Oliver Bilal, who leads the Europe, Middle East and Africa distribution business at Invesco. “So the majority of future clients will use digital solutions. AI will be the key to unlock the potential for serving customers in future.” Bilal has 23 years of experience in financial services, having worked at UBS and Natixis before joining Invesco.

But wealth management firms understand that many rich clients want the comfort of talking to a named individual; that Jim or Jemima is handling their portfolios. So far, AI has not resulted in job losses. Instead, where fund managers are deploying AI, it is being used as a tool to help their staff deal with clients. “Our aim is to improve client retention and client satisfaction,” says Aaron Bates, who leads the ultra-high net worth team at Bernstein Wealth Management, after working for the federal government in the office of the US Trade Representative. “AI allows us to focus on remaining competitive and be relevant for the next generation of wealth-holders.”

The allure of AI may send fund managers in different directions. BlackRock, the largest fund management group in the world, has set up BlackRock AI Labs although it has been tight-lipped about what is going on there. Other fund managers are more open about what they are doing.

At Sanlam Investment Management, Ford uses AI to help identify companies across all sectors that he may wish to invest in. Sanlam has a database of listed companies built using data compiled by stock exchanges and regulators such as the Financial Conduct Authority in the UK and the US Securities and Exchange Commission. Since companies face sanctions for misrepresentation, they are less likely to claim that they have a large AI-related business if they do not. Sanlam uses optical character recognition, natural language processing and translation platforms to pick out those companies that are benefiting from AI.

Just compiling this information is not enough, Ford says. The mere mention of AI in a corporate statement does not make it a suitable addition to the portfolio. Much depends on whether AI will have a significant positive impact on the business. Human skill is needed to put that information in context. “If I gave you a picture of Michelangelo’s ‘David’, along with tools and a block of marble, and said ‘see you in three weeks’, I would be disappointed with the results,” Ford adds.

A report by consultants Deloitte suggests fund managers will use AI to scan corporate profits statements to assess manager sentiment, analyse alternative data such as weather forecasts and container ship movements, and use corporate website traffic as an indicator of further business growth. In a sense, managers will be using AI as a more efficient search engine.

Fund managers can also use AI for fraud detection. Andreessen Horowitz says a large language model that has been trained on past reports of suspicious activity “should be able to identify a set of transactions that indicate a money-laundering scheme”. AI may also help investment companies to ensure that they keep on the right side of the regulatory authorities; fund managers face a host of compliance requirements from a range of regulators in different countries.

All these things will be very useful. Indeed, many service companies will be using AI for the same purposes; reducing the burden of bureaucracy or improving the efficiency of customer relations. Other financial companies will also be using AI for fraud detection, or for regulatory compliance.

There is no ‘piña colada’ system in which a fund manager can let the AI model loose and then go sit on the beach, sipping cocktails

There are some more sector-specific roles that AI can play. The Man Group is a fund management giant, best known for its quantitative-based investment approach. Gary Collier has recently become Man’s chief technology officer after a career in software development and technology management. He says the group has used machine learning to manage its trades. The approach uses reinforcement learning — a process that uses trial and error and rewards to train software to master complex tasks — to achieve better execution and reduce the market impact of orders. That is an important factor as the group trades a remarkable $7tn worth of assets in the course of a year.

Technology can help the firm in three ways, Collier says; it can help professionals innovate, help the group respond to change and it can make the firm more efficient. For that to happen, he says, the technology has to be put in the hands of professionals. As part of this strategy, the group has made ChatGPT technology accessible to the firm through a portal called Man GPT.

Bilal of Invesco says AI will also be useful in providing the specific products that clients want. Take the world of ethical investing, focused on ESG (environment, social and governance) factors. Many ESG funds are quite broad-brush in their scope but clients may have more nuanced principles; determined to avoid tobacco stocks but relaxed about alcohol, for example. AI can be much faster in searching the investable universe to deliver a portfolio tailored to the individual client’s principles.

In most cases, clients will probably not notice the effect of AI on their fund managers’ operations. Returns will not suddenly get better nor will their trusted advisers be replaced by a soulless machine, at least not at the top end of the market. Instead, investment companies will use AI to make their operations more efficient, and their interactions with customers more effective.

Will any cost savings be passed on to clients in the form of lower fees? It seems likely that the biggest influence in that regard will not be the rise of the robots but the remorseless forces of market competition led by low-cost index-trackers.

AI may end up turning the world upside down but when it comes to fund management, the changes will be incremental, not revolutionary.

This article is part of FT Wealth, a section providing in-depth coverage of philanthropy, entrepreneurs, family offices, as well as alternative and impact investment