9 Questions on Artificial Intelligence for Wealth Management
“I believe that by the end of the century, the use of words will have been altered so much that one will be able to speak of machines thinking without expecting to be contradicted.”
― Alan Turing, English mathematician, computer scientist and cryptologist, from
The foundation of AI started back in 1950 when scientist Alan Turing published a seminal paper containing a description of what is now referred to as the “Turing Test” which is designed to determine if a machine can think. A group of scientists got together at Dartmouth College a few years later and coined the term “artificial intelligence”.
Fast forward to this year and a panel discussion at the Invest 2017 Conference held in New York City on the ability of AI to revolutionize wealth management. The following industry experts were brought together:
- Jason Mars, CEO, Clinc
- Jeff McMillan, Chief Analytics and Data Officer, Morgan Stanley
- Joseph Kochansky, MD, Head of Aladdin Product Group, BlackRock
- Siddharth Sharma, CTO, Hedgeable
Implementing AI technology is an offensive move all the way, Clinc’s Mars stated. Technology companies have created high expectations and continue to raise the bar. But financial institutions are not technology companies. “It is a gnarly environment to navigate,” he explained.
The focus should be less on AI as AI, but more on how you are enriching people’s lives, Cinc insisted.
AI is much more offensive for Hedgeable, Sharma declared, but for large financial institutions it is defensive since they have to catch up to be competitive.
According to BlackRock’s Kochansky, development and integration of AI is inevitable and will become embedded in everything. BlackRock uses AI technology to understand portfolio construction and to measure risk across all asset classes. They are now taking the same technology used for their institutional portfolios into the wealth space, he confirmed.
Quality control is of ultimate importance when measuring risk, Kochansky emphasized, and AI has been instrumental in their efforts. Identifying the one investment that could melt down is like trying to find a needle in a haystack across trillions of dollars in assets. This is true whether it is for a few large scale institutional clients or for millions of accounts at wealth managers, he asserted.
There is an enormous opportunity sitting in front of us, Morgan Stanley’s McMillan noted, but the SEC and FINRA are catching up. The AI technology didn’t exist to drive this level of innovation until about three years ago. Now we are running into more business constraints.
There are some basic AI tools that every firm will need to survive, McMillan continued. Anomaly detection is one that everyone is doing because it cuts across many areas including cybersecurity, fraud, even identifying inappropriate employee online behavior. They are also leveraging machine learning to reduce client attrition. This has become a higher priority due to regulatory exposure, he insisted.
What are chatbot applications?
Chatbots are essentially command bots and can’t be engaged like another human, explained Clinc’s Mars. Clinc was spun out from Mars’ research on scalable AI at the University of Michigan. They have nine PhD’s on staff along with 35 post-grads all working to re-create a human experience in a chatbot, he stated.
Clinc’s chatbot was designed without pre-defined rules, Mars cautioned, which sets it apart from most other chatbots. Theirs learns like a child by leveraging a neural network, he added. Humans learn language in a very different way from how most AI systems work. Creating AI using static models is messy and does not enable unbounded conversations. Users can use slang when interacting with Clinc and the system understands it Mars asserted.
A future transformative moment will be when people will be able speak to Alexa like they would another human, Mars predicted. Only then it will impact our lives in a transformative way.
“Maybe the only significant difference between a really smart simulation and a human being was the noise they made when you punched them.”
― Sir Terry Pratchett, English author, from
Can AI Become More Human?
It was at this point in the discussion that two of the panelists started expressing a strong difference of opinion. McMillian insisted that a curated database was the best way to implement an AI system that can provide value to advisors.
Advisors just want standardized answers, McMillian insisted. They need the same rules that can properly respond to a 65 yr old retirees in Michigan as well as a 35 yr old married couple looking to buy their first house.
Such an AI system would contain all the curated intelligence contained in the firm stored in a structured way, McMillian proposed, and then advisors would be able to access the answers. This will be the differentiator in next 5-10 years, he predicted.
Clinc’s Mars disagreed. He pointed out that there have been structured databases available for a long time, but they will not be as effective as neural networks. He gave the example of calling a helpline and then trying to work with the structured system would be more of a burden. The power of neural network-based AI is that the user can ask questions in any way they want and the AI will understand, he announced.
McMillan insisted that the technology is not available yet to build a system as Mars described. There is no AI system available today that allows you to scan in unstructured text into a natural language processor (NLP), such as an article from the New York Times, and then have a user ask it 10 random questions about the subject matter, he stated. The only way to do this now is with human beings in combination with strong technology.
In a follow-up call about this topic, McMillan told me that he has tried to implement this functionality with five different AI technology vendors. But none of them could do it as they promised. He claimed to have already spoken to Cinc (although not to Mars directly) and learned they have taken a structured database and wrapped NLP around it. While it is a good system that can make complex and dispersed information more accessible, it can’t do what Mars claimed in this panel, he scoffed.
Mars defended his statement that his firm has this kind of technology available now.
“By far the greatest danger of Artificial Intelligence is that people conclude too early that they understand it.”
― Eliezer Yudkowsky, American AI researcher and writer
What is Deep Learning?
Deep learning software is designed to imitate the workings of the human brain. The software learns to recognize patterns in digital representations of data such as sounds and images. The more data that is “fed” into a deep learning system, the more accurate it gets, Sharma stressed.
In 2016, Google rolled out a word-embedding approach to deep learning, Mars described, and this technique has unleashed new possibilities. All English is mapped to numbers in a multi-dimensional space. The proximity of the words to each other is represented by numbers and this is used to train a neural network.
Amazon’s Alexa is an example of deep learning, Mars pointed out. The software that runs Alexa is a matching algorithm that constantly checks what you just said against everything it has learned from previous interactions with you.
Alexa is not a perfect solution, Sharma observed, but it’s an important first step. Hedgeable is exploring ways to use deep learning to enhance their customer experience, compliance and risk profiling. (See How Envestnet | Yodlee’s Startup Incubator Nurtures Data-Driven Innovators)
Can AI Increase Adoption of New Technology?
Kochansky believes that improving adoption of technology requires a firm-wide commitment to addressing advisor problems. Advisors do not need to be involved in every step of client management, he asserted. This is where technology can play an important role to handle the commoditized parts of the process.
However, when clients struggle with personal issues, they need an advisor, not a computer, Kochansky stressed. (See Advisor’s Guide To Choosing The Best Portfolio Rebalancing Software)
Can AI Improve Advisor Efficiency?
While some firms use AI to identify prospective clients and deliver advertising to them, McMillan insisted that marketing is actually the smallest piece of the success equation for wealth management firms. With over 16,000 advisors each with 100+ clients, Morgan Stanley advisors intelligent advice on how to efficiently engage with their clients. (See The Indispensable Advisor by Joe Duran)
One area that Morgan Stanley is applying AI and machine learning is to provide advisors with suggested activities based on analysis of client data and transactions. They are calling it Next Best Actions (NBA) and it is a massive effort to increase advisors’ capacity and efficiency without burdening them with the underlying technology.
McMillan described two of the functions the NBA is trying to automated. The first is product matching and the second is the client standard of care.
To match products, NBA looks at every client and analyzes all of their historic behavior and their defined goals/objectives. It then runs a matching engine to review the several thousands investment ideas produced across all the firm’s product groups and score every investment idea (0-100) for each client to rate whether or not they would do the transaction.
The advisors then have a list of which products would be most appealing to which clients and can contact just those clients instead of blasting emails to their entire book.
The NBA standard of care process ranks every client that an advisor should contact and what to contact them about. It provides advisors with insight to help them figure out who to engage for: 1) investments, 2) operations/admin (i.e. mandatory contributions), and 3) life events.
One interesting example of how Morgan Stanley is leveraging Big Data to improve the advisor-client relationship is around life events. They import data from the multiple listing services (MLS) that realtors use to find every house for sale over the last six months and then search for matches in each advisor’s book. The system then presents advisors with any matches along with a suggestion to contact the client.
NBA for advisors is like “having an assistant with 800 years of experience who went to Harvard and does everything you ask him to do,” McMillan declared.
How Important is Data for AI?
Effective AI requires starting with a data-centric company culture, Kochansky explained. Whether the data is coming from the CRM or the investment process, everyone needs to participate in a consortium to ensure the data is clean. There also should be a centralized platform, so that when one person fixes something, everyone benefits, he stressed. (See Can Big Data Make Risk Tolerance Questionnaires Obsolete?)
A lot of data is self-referencing, Kochansky noted. a risk measurement based on yesterday’s data
If you don’t have good data you’ll be in trouble, McMillan charged. NLP and algorithms won’t work without a strong underlying infrastructure. Unless you can make a compelling argument as to why the data is important, people won’t do it, he cautioned.
“Forget artificial intelligence – in the brave new world of big data, it’s artificial idiocy we should be looking out for.”
― Tom Chatfield, British technology theorist
What is The Impact of AI on On Robo-Advisors?
Robo-Advisors are based on concept that the work of advisors can be automated, Sharma noted. In the future, client engagement will be handled mainly through AI, he predicted.
One of the killer benefits of AI is enhancing and augmenting human abilities, but not in lieu of a human, Mars insisted. (See 3 Signs A Shakeup Is Coming to the Roboadvisor Market)
How Should We Think About AI?
AI and and all the fancy analytics that go with it do not mean anything unless they can help us to open up the business opportunities and industry challenges that we are facing, McMillan stated.
This is one area that McMillan and Mars agreed. Mars emphasized that we should avoid AI just for AI’s sake. We should look to create value from new technologies. “How can you improve the lives of your employees or customers?” he asked.
Kochansky suggested to start with the end problem and then define what you’re seeking. He predicted that soon, AI is going to be embedded in everything.
Out of all the ideas discussed on this panel, the Morgan Stanley NBA concept sounds like the future of wealth management to me. The technology offloads the grunt work and presents the advisor with a list of actionable ideas that is customized to their client base. AI tools will continue to evolve and eliminate more and more of the extraneous effort and allow advisors to spend more time on what they’re best at: talking to their clients.