Can Next Gen AI Upend Financial Services Over The Next Five Years?

The next generation AI revolution started with the launch of ChatGPT-3 in November 2022. Since then, venture capital activity has poured into the generative AI space and the tech giants have scrambled to respond. Alibaba has launched Tongyi Qianwen, Google
GOOG
has launched Bard, and Microsoft
MSFT
has launched Bing Chat (powered by OpenAI, the same organization behind ChatGPT).

Generative AI assistants like ChatGPT can write code, answer questions about all but the most niche topics, and can draft essays and emails. The technology is far from perfect and sometimes produces inaccurate answers. But on the assumption that generative AI technology will only continue to improve over time (an assumption that we’ll discuss more below), this breakthrough represents a profound turning point in the history of technology.

How will generative AI disrupt the stodgy financial services industry? In many respects, financial services has shown surprising resilience and resistance to the last 25+ years of disruption brought about by the internet and online services. For example, the United State’s Bureau of Labor Statistics estimated that there were 283,060 financial advisors and 445,540 insurance salesmen in the United States in 2022. Despite the rise of online competitors, Americans are still working with in-person (and often rather expensive) financial advisors and insurance agents.

In light of this strong resilience to online disruption, this article series will perform a deep dive focused on how next generation AIs will impact the financial services industry. This series will be divided into a Part One article and a Part Two article. Part One will focus on the impact of next generation AI on the industry in the short term. By the standards of the slow-moving financial services industry, five years is considered the short-term outlook. Part Two looks at how AI will change the financial services industry into the 2030s and beyond.

Before exploring how AI will disrupt financial services over the next five years, let’s reflect on the current state of the industry and why most financial services firms have yet to fully embrace AI. Recall that there was an earlier hype cycle around AI and chatbots in the 2010s. The enthusiasm was premature, and AI failed to live up to that earlier hype. The technology was simply not good enough for risk-averse financial services firms to invest the time and effort into embracing bold new approaches. As a result, the chatbot assistants that are currently offered by the financial services industry in 2023 are generally low quality. According to Andrew Way, Head of Financial Services Monitor research at Corporate Insight, “we have been tracking the chatbots offered by the industry in both retail accounts and workplace retirement plans for the last three years. With a few notable exceptions, the industry largely offers underwhelming AI assistants that struggle to answer complex questions and at best provide users with a link to a page to learn more.”

Now that the next generation of AI has arrived and the technology has taken a huge leap forward, could conversational AI assistants start to replace relatively high-paid white collar jobs such as financial advisors, insurance agents, bankers, loan officers, etc.? Not in the next five years. There are three broad factors that will prevent AIs as advanced as ChatGPT from massively displacing jobs in the financial services industry anytime soon.

First, the industry is constrained by onerous regulation

In most jurisdictions, the financial services industry is highly regulated. Generative AI’s impressive ability to produce reasonably accurate answers to most questions does not meet the very high bar set by regulations and legal liability. For example, imagine if a customer asked an AI a financial planning question that involves significant trade-offs. The customer could ask the chatbot something like “I just learned that I need to pay for my mother to move into a long-term care facility and I don’t have the funds on hand. Should I take out a home equity line of credit, dip into my kids’ college fund, or sell stocks that I bought a decade ago and trigger a large capital gains tax hit?” Any established bank, brokerage, or insurance firm would be extremely nervous to let an AI answer a question like that. If the AI answered incorrectly and misstated the tax implications of the various options, the firm could open itself up to lawsuits and/or regulatory fines.

While America is generally perceived as the most litigious major market, the regulation-heavy nature of financial services is not limited to America. According to Willem Röell, a Fintech lawyer at Amsterdam-based law firm De Roos, “EU financial services legislation contains strict requirements around what financial firms can and can’t do. In particular, the “black box” nature of AI presents many legal and compliance challenges. That said, the EU is starting to develop new regulatory frameworks for AI. The EU AI Act and the Dutch Central Bank’s SAFEST principals are examples of early developments in AI regulation.”

Second, the industry operates on a slow, relationship-focused sales cycle

Many financial services products are “sold, not bought.” For example, most consumers do not purchase a life insurance policy online or select a financial advisor after online research – they are sold an insurance policy or a financial advisor service through some kind of existing relationship or because they are contacted by a representative. In addition, financial services operates on a very slow sales cycle – consider how often you change your main bank account: not often. Even if OpenAI announced tomorrow that ChatGPT was pivoting to become a bank, it would take them many years to capture significant market share.

Consumer preferences will likely change over the long term – something we will explore in Part Two – but in the short term, the unique nature of sales in this industry limits the impact generative AI will have on jobs and employment.

Third, the industry cannot deploy next generation AIs until outdated tech stacks are modernized

The financial services industry is plagued by outdated technology infrastructure that makes it hard to implement a powerful AI assistant. Many financial services firms operate a messy, decades-old tech stack of different databases, complex vendor relationships, and outdated programming languages. Many firms have yet to complete the multi-year transition process necessary to operate a modern tech stack. For example, Citigroup
C
, the world’s seventh largest bank (excluding the four Chinese state-backed banks) stated in a Q2 2022 earnings call that they are currently working through a multi-year process to transition 16 bank ledger platforms deployed across 121 instances into a single cloud-based bank ledger.

Next generation AIs need to sit on top of a modern technology infrastructure. AIs need clean, consistent and well-organized data in order to be able to answer customer questions about their finances and spending history. According to Christos Ziakas, CTO of a stealth mode generative AI startup, “generative AI technology continues to improve and is getting better at collecting data from different sources of information. That said, the data has to be in a format that generative AI models can efficiently retrieve and interpret. The financial services industry also needs to modernize its AI infrastructure (such as tools for evaluating the performance of AI models) to support advanced generative AI solutions.”

Even if change is slow, next generation AIs are coming to financial services

Despite all of the limitations I have outlined above, next generation AI will inevitably change financial services. Slowly, but surely – it will happen. Firms will eventually work through technical and regulatory challenges. Five years from now, your financial services website and mobile app will be able to offer you an AI assistant that can answer far more advanced questions than current technology. Imagine being able to ask your AI assistant a question like “I want to take a vacation to Japan this summer. How much could I spend without impacting my financial goals?” The AI will be able to call upon knowledge of your financial plan and your average yearly spending across accounts to provide you with an instant answer.

In addition, over the next five years, firms will use the technology behind-the-scenes to make their employees more efficient. Some early pioneers started using AI to improve internal processes in the late 2010s. For example, Morgan Stanley
MS
has been using machine learning to improve financial advisor’s client interactions since at least 2017. Overall, however, the industry as a whole still has ample opportunity to make internal processes more efficient by using AI.

According to Laura Waldenstrom, a venture capitalist at Inkef Capital, “The application areas of AI within financial services go way beyond consumer-facing use cases. Internal operations within areas such as anti-money laundering, customer onboarding, and fraud protection can all be improved with AI. Historically, banks have been one of the slowest to adopt modern technology, but the recent developments around generative AI have only accelerated the sense of urgency among financial institutions to modernize their internal operations.”

In the longer term, there is potential for generative AI to start replacing financial services jobs in mass. Could financial advisors and insurance agents go the way of travel agents? Demographic change along with the rate at which the technology improves (remember that assumption discussed in the introduction?) could see AI replace many jobs in the industry. We will explore the long-term impact in Part Two.