"AI" appears to be on everyone's lips these days, and wealth management in North America is no exception. This article, by our US correspondent, takes views from a range of firms and advisors about artificial intelligence, including ideas on its benefits and limits.

Investment managers are approaching artificial intelligence with great interest and an abundance of caution.

Chief investment officers and portfolio managers at RIAs, family offices and wealth management firms are exploring and, in some cases, incorporating the latest advances in generative or conversational AI technology embodied by OpenAI and ChatGPT.

But artificial intelligence’s role in making investment decisions to date is limited, albeit burgeoning.

“We use ChatGPT as a productivity tool to summarize large research pieces as well as a tool to help in deriving executive summaries for our internal research reports,” said Matt Dmytryszyn, CIO of Telemus, a Michigan-based RIA. “Internally, our research reports can be detailed and lengthy. But not all audiences are interested in that level of depth, and having ChatGPT help to write the summary is a time-saving tool.”

The Colony Group in Boston is using ChatGPT and Bard, a chatbot developed by Google, to help develop code used for analyzing stocks and markets. The RIA, which has formed a committee of senior executives to study the use of AI, also works with a third-party vendor that develops quantitative models to rank individual stocks, according to CIO Brian Katz.

(For related comments about wealth management and AI, look here on the subject of behavioral finance, for example.)

Constraints: cost and expertise
At present, Colony is focused on using AI to “automate basic functions rather than manual analytical decision-making,” Katz said. 

Cost and expertise have been  the primary constraints, he explained: “Both problems are due to the fact that AI has not yet been widely adopted so there’s a real cost to using this technology at scale, and technical expertise is still limited in how we can use it more efficiently.”

Although Connecticut-based GYL Financial Synergies uses AI “sparsely” internally, the firm is accessing artificial intelligence through third-party providers such as Bloomberg and sell-side firms that use AI “to generate investment and trading ideas for their buy-side clients,” said CEO Gerald Goldberg.

GYL portfolio managers also use Microsoft’s Bing Open AI function and ChatGPT for stock research and coding aids, Goldberg said. One GYL advisor told Goldberg that he enjoys asking Bing questions like "I might ask analyst who has been covering a stock or an industry.”

Caveats
In these “early innings” of employing artificial intelligence, however, there are plenty of caveats that worry wealth management firms.

While AI’s ability to crunch and synthesize large amounts of data and recognize patterns is impressive, the technology’s reliability is suspect and “limited by the size and quality of the data base it has access to,” Scott MacKillop, CEO of Denver-based First Ascent Management, pointed out.

Artificial intelligence “sources information from the herd and is mostly dated and backward looking,” said Jon Ekoniak, managing partner of Bordeaux Wealth Advisors, explaining why the RIA doesn’t use AI to help make investment decisions.

What’s more, information on many of the funds the firm uses, particularly private and illiquid ones, are not readily available on current AI tools, Ekoniak added.

AI assistants like ChatGPT need to improve incorporating the most up-to-date news and information, agrees fintech consultant Grant Easterbrook. 

“Progress is being made on both of those fronts, but for a manager looking to make decisions with millions of dollars on the line, I don't think conversational AI is good enough just yet,” Easterbrook said.

AI assistants also have to “get better about being able to clearly explain their sources and how they arrived at a conclusion,” he said. “Currently, AI assistants can sometimes hallucinate and give wrong answers, and a manager would need to check that the information is correct.”

Artificial intelligence software are essentially programs “that are taking in data and sending out outputs,” Rob Santos, CEO of Los Angeles-based Arrowroot Family Office, pointed out. “For any numbers of reasons, the outputs may be incorrect or incomplete and certainly do not provide adequate disclosures or clarity on what aspects of the outputs may be unreliable either to a smaller data set or incorrect input data.”
 

‘Holy Grail’ or ubiquitous commodity?
Despite the current wariness surrounding AI, the technology “will be the holy grail of how people think about managing their assets,” Mary Callahan Erdoes, CEO of the Asset & Wealth Management division at JP Morgan, said at the bank’s Investor’s Day conference in May.

CIOs acknowledged progress, but remain guarded.

While AI can be used for analyzing stocks or managers quantitatively, Colony Group’s Katz questions “the sustainability of any competitive edge derived from AI as we expect it to garner widespread usage over the coming years.”

“If everyone is using similar AI technology,” he argued, “it might quickly arbitrage away any opportunities to add alpha. True differentiation may only come from the quality and uniqueness of the datasets used for the analysis.”

Artificial intelligence will improve the efficiency of investment analysis and portfolio management, but won’t make the final decisions, Telemus’s Dmytryszyn predicted.  

“Ultimately the analyst or portfolio manager “will need to provide their own judgment around the inputs,” Dmytryszyn predicted. “I think this will enable analysts to spend more time on higher-level, critical thinking functions. This should lead to wider perspectives on investments and be a positive.”

MacKillop expects AI’s reliability to get “better and better” and sees the technology playing a large role in research, comparisons and spotting patterns. AI will certainly advance “to the point where it can make buy and sell decisions,” he added.

But the question remains, "will they be good decisions?”