AI will replace robos to create new forms of customer value around advisory
By Chris Kapfer
Catherine Flax, chief executive officer of Pefin (US), a financial advisory firm that has built an enterprise wide artificial intelligence (AI) platform in the last six years, discusses the future of wealth advisory and why AI will beat robos and experienced relationship managers, without replacing the latter.
- Robos and relationship managers are not able to makes sense of the complexity of the interrelationship among all data that defines an individual’s financial life
- Improved data infrastructure and computational capacities are driving artificial intelligence (AI) applications
- Changing breed of AI-based wealth services take fiduciary advise and transparency to the next level
Robo platforms have seen adoption across all sectors of the financial institution (FI) industry since Betterment, the first stand-alone automated investing service launched in 2010. Fund offices, banks, mutual fund companies, asset managers, broker-dealers and a new breed of independent robo advisory firms followed aiming to run their services on a set of static rule-based algorithms. Global assets under management (AUM) by robo-advisors are estimated to reach more than $600 billion in 2017, according to UK-based Business Insider. In 2016, Fidelity (US), the world’s largest mutual fund launched “Fidelity Go”, a robo-platform that is aimed at younger audiences of 25 to 45 years old.
In the wake of the global financial crisis, the massive regulatory burden on wealth management businesses and a cut-throat competition in the high and ultra-high net-worth echelon of wealthy clients forced the industry to create new, cheaper and faster ways to acquire, on board and process clients, in particular appealing to the average investor or a younger emerging affluent segment that, often don’t qualify or can’t pay human advisors.
Automated robo services are focused on operating efficiencies, not advise. Its competitive value is a cost proposition by equally having a lower entry point of AUM. This is supported by a simple automated risk screening, a passive investment strategy with automated rebalancing, tax optimisation and data ownership vis-à-vis bank ownership of data. In the US, a typical robo advisor levies a fee of 0.25% per annum compared to 1% to 1.5% charged by a human advisor.
Recently, Royal Bank of Scotland (UK), under its Natwest brand, launched a robo-advice proposition, targeting its five million customers. The service, which will be available to clients at a $672 minimum investment with a $13.44 process fee, aims to plug “the advice gap”, with the total average cost of advice, investment and platform fees set at 0.95%.
“Robo advisors have been fairly innovative in some respects. For instance, they ensure a very smooth online on boarding experience- particularly for an industry stuck in 1970s technologies and 1940s regulations. The challenge for robos has been to fully capitalise on this innovation, to build and innovate further and provide a new level of scalable insight and depth of service, which could provide and translate more complex investment choices, highly customised to personal needs. On this they have unequivocally failed,” according to Catherine Flax, chief executive officer of Pefin, an AI financial platform focusing on financial planning and advice.
According to Flax, robos have disappointed in three key areas: providing advice that people need, providing better investment choices and platforms and lastly, personalising and linking an investor’s actual goals and financial needs to investment recommendations.
The combination of these three factors have resulted in weak product adoption and higher customer acquisition costs, leaving consumers with the perception that technology won’t be able to be the “game changer” in finance, according to Flax.
“One of the key concerns we have with robo advisory platforms is that they don’t have a complete picture of the person they intend to serve and thus are unable to make suggestions about what appropriate investments should be. Robos typically don’t adapt their investment allocation strategies and many use simplistic investment models based solely on the Modern Portfolio Theory conceived in the 1950’s. From our perspective, what is fundamental is making sure that people are getting the right advice for what they are trying to achieve, and in many instances, the right answer for a person may even be that they shouldn’t be investing.”
“Robo advisory is still quite nascent. There are in the order of $20 billion AUM with robo advisors today in the US. If you compare that with the trillions of dollars of AUM with the major traditional advisors, it’s about $36 trillion in the US alone. There is a tremendous opportunity across a lot of different client segments.”
The robo advisors have particularly targeted younger people who have never invested before. It is an important market, but the biggest risk robo advisors face today is a significant market downturn - and for those of us who have been in financial markets for years, it’s not a question of if there will be a downturn, but when. Because Robos are giving generic advice and there is no accompanying financial literacy, if you have a segment of people who don’t understand that markets are volatile, and don’t understand the risks they are incurring by investing. When there is a downturn, my expectation would be that there will be a dramatic reduction in AUM from the robos as people pull their money out.
Robo advisory companies increasingly struggle to offer 100% automated services only. Betterment conceded early 2017 that it will offer access to human financial advisors charging a 0.5% fee for $250,000 minimum balance and unlimited access to financial advisors vis-à-vis its pure digital plan at 0.25% per annum. Increasingly, FIs are blending robo services and live advise, but risking to turn those propositions into a human relationship management approach, which was broken in the first place. Bringing back the human advisor in the picture defeats the very aim robo advisory platforms were set out to achieve – scale.
“It’s the worst of both worlds (the hybrid model- robos with humans, people can call). By having human advisors, they aren’t collecting better or richer information about the client. They still operate in a relative vacuum regarding the information they should have about their client. You are not necessarily getting the same advisor when you call, contrasting with a private bank model where you have the same advisor and where you develop a more trusted relationship over time.”
“There is also one often overlooked and potentially more serious issue with this hybrid model - it still relies on the individual who is calling in to come up with the right questions to ask the advisor. The onus is thus being put on the client to know what they should be asking. And to me, that is a fatal flaw in the model.”
“Because the human advisors hired by robos provide a voice on the phone, it feels more like advice, but the analysis hasn’t really improved beyond the robo, hence the advice is neither good nor fiduciary,” she added.
Robo platforms will not survive in its current stage until firmly focused on creating new forms of customer value around advisory. Hence, they will only serve as an intermediary stage in the evolution of the wealth industry. In fact, the largest global commercial banks and independent broker-dealers, such as Morgan Stanley, LPL and Citibank, are moving away from robo advisory into AI and its associated “cognitive” tools to enhance client relationships.
“The difference is AI - a combination of dynamic algorithms, neural networks, reinforcement learning, as well as generalised AI, that traditional banking models and robo advisors don’t use. AI is able to assess the complexity of the relationship among all the data points that define someone’s financial life such as income, savings, spending patterns, debt and investments that are meshed with your life decisions and the wider macro-economic environment that is important to a client’s financial life and the markets. If you look at the intersection of these relationships, a human being doesn’t have the computational capacity to analyse them. Robo advisors don’t have the technology as they are based on simplistic models, not analysing individual spending patterns. Imagine, at any moment of time there are thousands of relationships, and if you propagate that forward on a moment by moment basis, by up to 80 years, you are talking about millions of data points and interrelationships. These data points include the national tax code, local taxes, social security rules, Medicare - a single-payer, national social insurance program in the US for the elderly - and individual savings patterns.”
“In addition, plans that come from traditional human advisors tend to be static. By contrast, AI platforms continuously update plans, providing information to people as and when anything such as tax code, spending patterns or market information changes. A human advisor is looking to update on a quarterly basis at best. They would meet with a client once or twice a year and the fees associated with it would often be larger than anyone can reasonably anticipate making in those investments,” said Flax.
Traditionally, banks found it almost impossible to separate the role of advisor from the delivery of the product – making it impossible to stay impartial. With AI, those two functions will start differentiating themselves in the advisory industry. Pelfin claims that it provides customers the choice to execute the advice at any time elsewhere.
“This is a paradigm shift that I expect is not going to only take place in financial services, but more broadly. There are ways in which trust can be built – and one way is to separate the advice from selling a product. Traditional financial institutions have to figure out how to they reconcile fiduciary advice versus pushing a product,” she added.
While there has been generalised AI platforms in the market such as Watson, those are not suitable for specific industries because the system has to learn behaviourally over time in a specific subject matter. Enterprise level platforms developed for one purpose take time to build, hence there are much less AI platforms in the market then robo advisors. Pelfin has been building its AI platform in house over the least six years.
“AI is a broad field. There isn’t just one type of AI - and there is range of how cognitively developed the current tools are. However, the advances in the last ten years have been astronomical – allowing for a much richer ability of algorithms to understand and interpret data. This makes it possible to go beyond the generalised - such as a computer being fed a million pictures of a cat and then being able to discern when shown a cat or a dog - to the very specific. It is the specific AI that allows for innovations like self-driving cars, or being able to interpret medical results for a patient - or to be able to provide unique and tailored financial advice to an individual.”
According to Flax, an increasingly available global data infrastructure, a new generation of computational power and dynamic algorithms are prerequisites for AI to build consumer facing applications. Natural language processing is the next big thing in the AI industry.
“If you talked to Siri or Alexa they are pretty good but they are not yet perfect. The Amazons and Googles of the world are really well positioned to continue making those developments because all of that requires vast amounts of data which they have access to, and so, the fact they are making those platforms open source is a big step forward. Companies like ours will build on top what they develop and the more people build on it, the better it will get,” she said.
Still, one of the big issues in AI is as they employ complex neural networks that even their scientist cannot explain how they arrive at the answers, decisions are perceived delivered from a “black box” and must essentially be taken on trust. How one ensures that a decision can be diagnosed backwards through the program’s neural network to reveal how a decision was made?
“It is very important that the results of the AI algorithms be validated, and that the “black box” element be removed. This is essential whether it is for financial advice, medical advice, or to be sure that your self-driving car takes you to your destination safely. With Pefin, we have ongoing analysis of the results – and validate these results with experts in accounting, actuarial science, finance and more. It is very important that our clients understand why they are getting the advice that they are getting. Of course, some people aren’t interested in the details, but for those who are, we allow them to click in to see why advice is changing – was it a change in tax code, was it a change in your spending patterns or something else? Whatever it is, we are able to take users back to the root of the logic, which is very important. Irrespective of how the advice is derived, a client needs to understand their finances and should be able to trace the factors determining the net result to a cash flow/income statement/balance sheet driven model,” said Flax.
Keywords: Pefin, Robo-advisory, AI