We hear much commentary today about the use of AI "co-pilots" in financial services. What do they do, and how far will their use go? This article is part of a series in which we examine use cases for AI.

In the blizzard of stories and comments about how AI is burrowing itself into wealth management and financial services, a term that comes up regularly is that of the “co-pilot.” 

This aeronautical term is becoming more common. It conveys the idea that AI can “sit” alongside a banker, analyst or RM to perform work such as identifying investment ideas, spotting data “red flags” and handling compliance chores. 

Many of the steps for advisors at a private bank and wealth management firm involve pulling together disparate information, often supported by different tech stacks. This then has to be synthesised to generate insights. Generative AI can quickly process all this, saving advisors and others hours of time. And time is money. 

As explained in an article on this news service back in 2023 by F-Prime Capital, AI has enabled these co-pilots to emerge and handle routine tasks such as reviewing legal documents, opening accounts, preparing client presentations, adjusting asset allocation, requesting query service, addressing ad hoc questions, and other activities beyond their core role of advising clients, which currently takes up 36 per cent of advisors’ time. The average advisor spends more than two hours “behind the scenes” for every hour they spend with clients.

So far, the main benefits in this context are “productivity enhancement, saving time and costs,” SimCorp’s chief product officer Marc Schröter told this publication. “That is what we see everyone working on, both on the wealth management side and among vendors.”

Schröter spoke to this news service as it is continuing to explore AI use cases in wealth management (see here and here and here for examples of our articles). 

The co-pilot image is different from the idea of AI replacing human advisors. Even so, automating parts of the financial sector value chain will probably lead to some jobs vanishing. Global banks will cut as many as 200,000 jobs in the next three to five years as artificial intelligence encroaches on tasks currently carried out by human workers, according to an analysis by Bloomberg Intelligence in January this year. Back office, middle office and operations are likely to be most at risk, while customer services could see changes as bots manage client functions, it said. On the upside, changes could boost banks’ earnings. In 2027, banks could see pre-tax profits becoming 12 per cent to 17 per cent higher than they would otherwise have been – adding as much as $180 billion to their combined bottom line – as AI powers an increase in productivity. That, at least, is the hope. 

While the noise level around co-pilots is rising, there's still a gap between that talk and what's actually happening. According to Ireland-headquartered compliance technology firm Fenergo, only 1 per cent of the banks which it has surveyed successfully automated the majority of their KYC and onboarding workflows; there is a growing appetite for AI-driven solutions. Some 38 per cent of respondents indicated plans to deploy AI to enhance operational efficiency, while 30 per cent aim to improve data accuracy with AI-powered tools. There is work to be done. Recent SimCorp research among 200 senior leaders at asset managers and asset owners, showed that 75 per cent of them said they were “somewhat prepared” for AI. There is still a great deal of experimentation with deciding what are the most suitable and credible use cases.

Examples of AI in action
Schröter at SimCorp said one typical use case predicting settlement failures based on the financial instrument being traded, the specific counterparty, and the clearing broker. In portfolio management, an AI co-pilot can examine portfolio behaviour; it can see what parts of it exhibit specific risks, where the risks appear to be at their highest and most at variance with stated goals, and other points. “It is faster and it is easier,” he said. 

“Another example for a co-pilot function is helping a portfolio manager with data queries about their portfolios. This could be `what are my top 10 holdings?’ It could be understanding the portfolio’s currency exposures, what has contributed most to returns year-to-date, and so on. The key is to have access to the necessary data,” Schröter said.

AI can help flag important, impending events and remind managers of them, Schröter explained. 

Schröter gave examples of actual AI uses such as how it works in SimCorp’s Axioma risk and portfolio optimisation offering. The Axioma wealth management solution WealthLens helps wealth managers identify which – out of potentially thousands of client accounts – might need to be rebalanced. The solution uses AI to analyse past behaviour to determine the probability for a portfolio needing to be rebalanced to meet its investment mandate. Ultimately, the human retains the decision-making control, not the AI.

In private equity, machine learning and AI can automate the processing of data from important documents, such as capital calls and distribution notices, which investors often receive in various forms and formats. By automating data gathering, investment professionals can spend more time on value-adding activities rather than manually copy-pasting rudimentary information.

The possibilities, it seems, are endless. The challenge for AI solutions and use cases, Schröter said, is when a firm's data is not connected but is in silos. “There are lots of people [at SimCorp] looking at how to create AI solutions based on our centralised platform SimCorp One,” he said. 

The payoff
Where, ultimately, does the “rubber hit the road” in all this for the clients?

“The main driver of all this [AI use case] is cost-pressure,” Schröter replied. 

Clients will see benefits in terms of less upside pressure on costs, as well as having more tools to ask about performance, fees, risk, and improve engagement with firms serving them, he said. Looking ahead, AI will become more autonomous, with “co-pilots” not always needing to be prompted to provide information and flag up issues, Schröter added. 

At which point, the co-pilot is going to do a lot more of the flying.