Private banks believe new technology has the ability to deepen client relationships, though stringent measures must be in place to ensure responsible use
The launch of artificial intelligence (AI) chatbot ChatGPT by US AI research laboratory OpenAI last November has opened new tech frontiers in wealth management, leading innovative players to launch new services based on large language models (LLMs) and generative AI to empower advisers and drive greater client engagement.
This technology, which can generate or create new content based on a variety of inputs, has polarised opinions, with many highlighting huge risks including a potential threat to humanity. But there is no doubt it can offer significant benefits, if used with appropriate controls and on trusted sources.
While JPMorgan Chase is currently developing a ChatGPT-like software service that leans on a disruptive form of AI to analyse and select securities tailored to customer needs, earlier this spring Morgan Stanley Wealth Management (MSWM) launched a service leveraging OpenAI technology and the bank’s intellectual capital. The bank’s aim is to deliver content and insights into companies, sectors, asset classes, capital markets and regions around the world into the hands of financial advisers, in seconds.
“Data shows that financial advisers drive the most value for clients when they engage with them, and our goal is to use generative AI to help broaden and deepen client relationships,” says Jeff McMillan, head of analytics, data and innovation for MSWM. Financial advisers and their teams can use this internal capability to ask questions and contemplate large amounts of content and data, with answers generated exclusively from MSWM content and with links to source documents.
“The idea is that we give our clients access to the most expert individual on any given topic, 24 hours a day, seven days a week, instantaneously, empowering our financial advisers to help their clients and add more value,” he says. “It will be like having our chief investment strategist, chief global economist, and global equities strategist on call for every financial adviser 24/7.”
The US wealth management firm had already launched AI projects, including an internally built AI-based engine, called Next Best Action, delivering “timely” messages to clients and prospects, guided by financial advisers.
“But Open AI is a quantum leap forward in terms of its ability to understand, consume, and then play back content,” says Mr McMillan. With LLMs, the ability to ingest any type of content, consume it, and then make it available – i.e. the time to market – “is infinitely smaller”, and the ability to support multiple types of content is “infinitely greater”, he says, with accuracy also “much better” than with traditional content retrieval.
While everyone can access OpenAI’s LLMs, firms need to have two key capabilities to be successful in this space, says Mr McMillan.
“First, you need unique content, as there is no real artificial intelligence without real intelligence. These machines don't just make things up. They require deep intellectual capital to drive knowledge and capabilities.”
Also, technology is very complicated. “It is not like a refrigerator, which you plug into an outlet and it works. We have had to spend a lot of time developing that skill set internally,” he says. “It is crucial to be able to fine tune these models to make them work on your data, your content, which is a complex process that takes months of effort,” he says, adding Morgan Stanley is one of the few firms working directly with Open AI.
Importantly, data needs to be structured and tagged appropriately. Over the past five years, the US institution had already built a centralised content repository with all data, and a tagging facility for the purpose of general search, which had to be readjusted over the recent months to launch this new service.
“Where we have had to spend a lot of time is understanding how to tag the content, to help the machine understand when a piece of content is relevant enough,” says Mr McMillan, highlighting the importance of content recency. “In many cases, we've had to change our tagging to help the machine do a better job.”
Intentionally, as with any new disruptive projects, the bank does not use any client data in the model, because of “issues and risks around it”.
Merrill Lynch is also exploring how to use LLMs to support advisers to service clients and leads, to help them grow their business and create efficiency, says the firm’s Nitesh Kadakia, head of innovation and adviser platforms.
The US bank has been “constantly investing in AI technology” for several years now, and today uses AI primarily in the areas of proactive client engagement and servicing. A natural language search capability app for advisers and clients, Ask Merrill, powered by the bank’s own AI engine, Erica, leverages technologies in advanced analytics and cognitive messaging.
Today, the institution is looking both for partnership opportunities and to build internal capability in AI, with the outcome likely to be a combination of both, reveals Mr Kadakia.
Only a handful of players, the OpenAIs and Googles of this world, will be able to afford to build LLMs, but hundreds of smaller firms will develop a myriad of interesting use cases and may partner with these large tech firms or with big banks, he predicts.
Through LLMs, wealth managers will be able to personalise the marketing collateral for clients, explain complex information, such as regulation or investment research, simplifying financial language. Pulling information across different data sources, this technology will automate the delivery of financial opportunity for advisers, allowing them to be more proactive in reaching out to clients, also allowing them to deliver more “family office type service offerings” as they are occurring, for instance around healthcare or college planning.
“What is important is to prioritise the right use case, identifying business drivers within any organisation, and use the technology to differentiate the business from competition,” says Mr Kadakia.
Trusted sources
Analysing which use cases third party tech partners can offer is therefore crucial in the selection process, with preference going to vendors able to gather information from trusted sources that align to the bank’s values, and that have “a large volume of content”.
“The most important consideration about LLMs is that technology must be applied across trusted sources, and information needs to be accurate and up to date, to deliver timely, consistent advice to advisers,” says Mr Kadakia. Merrill is exploring the opportunity to apply LLMs not just on its data sources but also on trusted government sources.
Indeed, one of the biggest drawbacks of ChatGPT is that it cannot be used as an authoritative source of information, as it still relies on content from the internet taken in 2021, which is filled with misinformation and may not be up to date.
At DBS, the Singapore-based bank, AI and machine learning is used “extensively”, for instance to send personalised nudges directly to clients’ email inboxes or phones, or to empower relationship managers with prescribed talking points to have “meaningful” conversations with their clients.
Today, the institution is “in the early stages” of developing “seamless integration” of generative AI technologies “to produce high-quality, personalised advice at scale”, enabling advisers to respond proactively to evolving circumstances. “Our goal is to magnify the scale, timeliness and richness of insights we deliver to our clients,” says DBS Bank’s group head of cognitive banking and martech, Royce Teo.
We firmly believe that AI and large language models will play a pivotal role in shaping the future of private banking.
As the complexity and pace of financial markets increase, the need to rapidly process and act on vast amounts of information is paramount, he says. “AI and large language models can analyse extensive datasets, discern patterns and generate insights at a scale that far surpasses human capabilities. This allows us to offer more personalised, timely, and effective services,” believes Mr Teo.
If the application of AI in private banking offers key benefits, including productivity gains and improved customer experience, two significant risks need careful management, says Mr Teo. These include data privacy and ‘model explainability’.
“With AI requiring substantial personal data, stringent measures must be in place to ensure data protection and responsible use. Furthermore, to avoid biased or discriminatory decisions, AI systems must be trained on diverse, representative data and regularly assessed for bias. Balancing these benefits and risks is crucial to fully harness AI's potential in private banking,” believes Mr Teo.
HSBC is also “closely looking at AI, vigilantly exploring and evaluating the benefits of adopting generative AI for our staff and clients,” explains Anil Venuturupalli, chief operating officer and head of digital transformation at HSBC Global Private Banking and Wealth.
“While human interaction remains at the core of service offering, private banks would benefit from AI capabilities that support client-facing staff to spend more time with clients,” he says, acknowledging that generative AI models often need “millions, if not billions, of data points to work effectively”.
As with all disruptive technologies, it requires frequent training and monitoring to ensure the technology is achieving its intended outcomes and benefits to enhance client and staff experience, he warns.
“To fully leverage LLM capabilities, we first need to assess the right use cases that will help our advisers provide better service to our clients.” RMs and advisers should also be trained to identify data biases and “exercise careful judgment when using outputs from generative AI models”, he adds.
There is little doubt that AI and LLMs have the potential to have a profound impact on the industry, says Doug Fritz, co-founder and CEO of F2 Strategy, a San Francisco-based wealthtech management consulting firm. However, as with any nascent technology, it is premature to predict the evolution and regulatory response to such development, he believes.
“As the relationship service model continues to be a core consideration for clients, we should consider AI to mean ‘augmented intelligence’ rather than ‘artificial intelligence’ for the private banking industry.” AI cannot replace the human relationships and advice provided to clients today, but it can empower staff with greater knowledge and intelligence, so they can be more responsive and proactive in attending to client needs, he says.
But to add value, generative AI must be used on data that is structured, reliable and up to date, and only a small percentage of wealth management firms globally have access to their own data in a structured manner, including CRM and historical transactional data, he says.
This means the industry will be split between “the haves”, large, global private banks which have their own data scientists and machine learning teams in their own organisation and have already been working with AI, and the “have nots”. The latter are generally smaller wealth management firms without the means to invest in technology.
The true differentiating factor does not merely lie in the adoption of these technologies, but in the way institutions utilise them to elevate client experiences, improve decision-making processes, and drive operational efficiency, says DBS’s Mr Teo.
“Ultimately, the most impactful use of AI and large language models will be to complement and amplify the human touch in wealth management, enhancing our ability to deliver superior value to our clients in a dynamic financial landscape.”
