“Any sufficiently advanced technology is indistinguishable from magic.”
— Arthur C. Clarke
This rings especially true in the rapidly evolving landscape of financial advisor technology. Over the next five years, significant changes are expected: streamlined systems, the impactful role of generative AI, and the evolving advisor-client relationship.
Recent findings from PwC’s 2023 Global Asset and Wealth Management Survey underscore this transformation. The survey predicts a surge in assets managed by AI-enabled robo-advisors, reaching nearly $6 trillion by 2027, almost doubling the figure from 2022. This trend extends beyond robo-advice, with more wealth management firms experimenting with generative AI for enhancing trading strategies and analyzing unstructured data.
As AI overturns the technology tables across industries, several pivotal questions are emerging:
- Will advisors have access to a more unified and streamlined system?
- How will the rise of generative AI transform the functionality of chatbots?
- And will these technological advancements narrow or widen the gap between advisors and their clients?
The wealth management industry is witnessing a tectonic shift towards more integrated and sophisticated software solutions. In a discussion with Dave Christensen, CEO of digital marketing provider FMG, We explored various emerging trends, including the impact of their recent acquisition of MyRepChat, the evolving role of chatbots, and the broader implications of AI in wealth management. (See It’s Alive!: Unleashing the Power of AI in Wealth Management)
Advocating for an all-in-one solution for tech stacks in advisor firms has been a consistent theme in my work. As Christensen explained, FMG’s approach to expanding their platform both vertically to address a firm’s diverse needs and horizontally across industries aligns with this philosophy. Their insurance division, Agency Revolution, is an example of their cross-industry expansion. Their strategy includes acquiring businesses in their target expansion sectors, while ultimately, everything operates on a unified platform.
This holistic approach ensures that insights gained in one sector can enhance another. However, one area of divergence is in communication, particularly texting, due to the different compliance requirements in wealth management. (See Reaching The Apex of AI Applications for Advisors)
Texting as a mode of communication in advisory settings is clearly underutilized, mostly for fear of running afoul of regulations. Recent stats show that text messages have a 98% open rate [CTIA, 2022] and 85% of people check a text within 5 minutes of receiving it. [SimpleTexting, 2022]
There is tremendous growth potential in compliant texting and its integration into other tools, yet many RIAs have not deployed these tools into their ecosystems.
Seeing this trends led to FMG’s recent acquisition of MyRepChat to cater to the texting needs of their wealth management clients, Christensen said. MyRepChat was the only independent provider of compliant texting software that competed with Smarsh, MessageWatcher, and Orion’s Redtail Speak.
The acquisition feels like a smart move since texting was a communications channel that was a gap in FMG’s marketing product suite that includes advisor websites, email, social media, and events.
Financial planning guru Michael Kitces said this about the MyRepChat acquisition: “And given that arguably the biggest challenge other than compliance obligations that advisors have around texting is how to fit it into their existing marketing processes, FMG seems uniquely positioned to integrate MyRepChat’s texting capabilities as part of an all-around marketing strategy rather than one-off responses to client questions.”
Regional communication preferences also play a role. In some parts of the world, such as Singapore, extensive information is often relayed through messaging apps like WhatsApp, a practice less common in the U.S. This highlights the varied approaches to communication in different markets and the potential for adopting new methods in the U.S. (See Ep. 144: Artificial Intelligence Powered CRM with Brian McLaughlin)
Our conversation then turned to chatbots and their limitations. Currently, the experience with chatbots is often transparent and limited, leading users to disengage quickly. My aspiration is to see chatbot functionality so seamlessly integrated into client interactions that the transition between AI and human communication becomes imperceptible. Studies, like the one conducted by Morningstar with their chatbot, named Mo, demonstrate that even minor enhancements like adding an avatar can significantly deepen client engagement.
Their chatbot was built on a new internal platform called the Morningstar Intelligence Engine. When Mo receives a user’s question, the engine identifies the most responsive content to provide to OpenAI’s large language model to construct a response. The response is then tested for relevance and responsiveness, before being fed back to the user.
According to Morningstar, they chose to work with Microsoft’s Azure OpenAI Service because of Microsoft’s commitment not to allow user data to train the OpenAI large language model. In keeping with Morningstar’s commitment to privacy, users are instructed not to input personal or confidential information into Mo.
Chatbots still have an upper limit on their usefulness as anyone who has interacted with one on an airline or cable company website knows. I know that I try to get to a human right away. “Agent” is most often my first message to any chatbot unless I have an incredibly simple request. (See AI’s Judgment Day: How ChatGPT-4 is Reshaping Wealth Management)
Integrating generative AI is a big step towards improving chatbot ability to communicate about complex topics with the goal being to handle the majority of calls that currently require human intervention.
Adding an avatar to a chatbot can provide tremendous benefits over text-based chats. A chatbot’s influence via emotional expressions on service outcomes has not received much attention.
Recent research has shown that human-like avatars that have the ability to display expressions of concern can increase customer satisfaction by reducing expectancy violations. “In particular, customer’s goal orientation, the human-likeness of chatbot’s avatars, and the relationship type between customers and chatbots can moderate the negative relationship between emotional expression and expectancy violation. These findings advance research on the emotional expressions of chatbots and provide critical insights for deploying chatbots in customer service…”
The journey to creating effective generative AI tools is complex, especially given the limitations in AI’s contextual understanding. At the heart of every generative AI application is a large language model (LLM). An LLM is a machine learning (ML) model trained on a large body of content—such as all the content accessible on the internet.
At the core of every LLM is something called a vector database. A vector is a mathematics term that is used in computer science as a representation of multidimensional data. In an LLM, vector data stores measures of proximity between words to enable the generative AI to understand the contextual relationship. (See The Journey Towards Data-Driven Wealth Management)
Christensen shared that FMG was an early adopter of vector databases and first used them for customer service to analyze data to help their AI understand technical concepts and maintain a long-term memory. This allows the LLM to assist when the team is executing complex tasks.
FMG decided to take all of the knowledge they have inside their company and feed it into their generative AI tool, Christensen explained. “For the past few months, every email, text, or in-app chat interaction with one of our service is going through this system and being refined,” he said. They’re close to a point where AI can accurately respond to customer queries 99% of the time, at which point they’ll just let their customers interact with it directly inside the application.
This type of generative AI model is particularly potent in the property and casualty insurance sector, where a tough market is resulting in huge increase in customer service requests. On that side of the business, FMG can access policy information through the Agency Management System– which is similar to a CRM but for the insurance business– and feed it into a generative AI which can then give customers information about their own policies. (See How Advisors Can Avoid Being Buried Alive by Technology)
Christensen is adamant that the industry must prepare for a future dominated by chat-based software. One significant benefit of this shift is the speed of development and response. FMG’s use of natural language processing to parse and understand client documents and interact through APIs is a prime example of leveraging new technology to streamline processes and improve efficiency.
One of the main advantages to a chat-based development structure is speed. “When you’re trying to build a product, you always have way more ideas than you have resources to execute as fast as you want,” Christensen explained. The R&D group at FMG is attempting to speed things by building repeatable processes based on AI.
For example, they trained an AI model using their internal API documents so that it could build code that interacts with their APIs based on a series of questions. This allows the web developer to put a chat over the API for editing a web page on their platform. A user could then ask the system to add a team member to the page, and the system would respond by requesting all the information and links it needed. It would then update the team member information on the website as instructed. (See Delivering More – and Better – Business Insights Through Predictive Analytics)
Christensen said that they have about 25 process templates with the most popular being one that can generate product specification documents. The template prompts the user to describe what the system needs to do. The system uses the template to generate the entire specification document in the proper format. It requires 10% of the time the company used to spend to write specifications, he reported.
If the process became standardized, it would be possible to build applications that are API-first, where everything is created after the API to fit it. FMG has their own interface over the Open AI API, which allows them to create their own templates for repeatable processes and their own system prompts for different use cases, and then share them throughout the organization.
Generative AI is also finding applications in tasks such as content creation, social media management, and developing interview questions. Many firms in the fintech sector are exploring the integration of generative AI into their systems to enhance client communication and automate workflows.
Pulse360 is one wealthtech provider that saw the benefits of generative AI before the rest of the world. They launched a new feature called Rephrase in September 2022, a full two months before OpenAI released ChatGPT. Rephrase enabled advisors to quickly summarize long blocks of text or entire articles or expand on a bullet list. Both of these capabilities are now table stakes for any genAI tool.
Additionally, there are innovative AI applications like Pi.Ai being used for brainstorming and strategy development, utilizing different models than those used in content generation.
Staying on the cutting edge of technology means imagining a future one step beyond what most people can see. For wealthtech vendors like FMG, it means expanding their platform by leveraging AI and chatbot technology. In an industry that has long struggled to embrace technological change, the providers that are able to harness generational AI to improve advisor productivity will reap the benefits of deeper client relationships and increased market share in the future.