What’s real — and what’s hype — behind AI

To find out what’s real — and what’s hype — behind machine learning and AI, try asking Siri two questions. First, “What is the average size of an elephant?” Siri’s machine learning technology has been trained with enough data and questions along the same lines to pull up the answer. (My Siri told me African elephant bulls are between 9 and 13 feet tall.)

Now, ask Siri a second question. “How long would it take an elephant to swim from Antarctica to Mongolia?” Stumped.

But, if you ask an eight-year-old that same question, she would tell you it’s impossible. Elephants can’t swim across an ocean and certainly don’t live in Antarctica. And, if that elementary schooler paid attention in geography class, she would also point out that Mongolia is one of the most landlocked countries in the world — 400-plus miles from the nearest ocean. In short, machine learning can learn how to complete tasks with enough data. But the technology has a much tougher time with putting information into context.

Banks are ahead of the curve when it comes to adopting machine learning-based chatbots because retail banking deals with a lot of repetitive questions. For them, there’s nothing “hype” about the advanced digital tools. For example, questions around how long it takes to open a checking account and how to access statements online. Chatbots can also follow up with customers by text to let them know, for example, that they’re over their target entertainment budget for the month.

Chatbots have certainly proliferated. Bank of America’s Erica, Wells Fargo’s Facebook bot, Fidelity’s Virtual Assistant, HSBC’s Amy, Capital One’s Eno, Commonwealth Bank’s Ceba, and USAA’s mobile app chatbot. Other financial services verticals have seen more limited adoption of chatbots. There are chatbots for saving for 401(k)s, college, real estate, debt collection, insurance, and more. A company I co-founded, Dream Forward, is one such example of an industry-specific chatbot. In our case, it’s focused on retirement planning.

There are also a variety of machine learning companies trying to help large organizations monitor things like CRM, payroll, invoices, security, fraud and more. The financial services industry is uniquely appealing to these machine learning companies for two reasons. First, compliance, money laundering, paperwork and fraud are a larger burden relative to other industries. Second, compared to other industries, financial services have an incredible amount of data on their customers.

Given these unique challenges, there are a variety of new tech firms selling financial services firms machine learning capabilities for compliance, fraud detection, and custom profiling for upselling opportunities. Some of the larger firms have also already started to tackle these challenges with their own technology. One of the best known examples of this is JP Morgan’s announcement that it had used machine learning to automate document review and saved 360,000 lawyer man-hours.

This is the area of financial services that is the hardest to unpack. Only select groups of employees at large financial services firms know the true extent to which machine learning with no human oversight is being implemented. There’s a lot of potential for hype that doesn’t match up with reality, or for marketing departments to exaggerate the uses of AI in press releases to drum up interest. Just because big firms are talking about AI doesn’t mean they’re ready to take on the risks associated with handing over tasks to an AI or that they’re budgeting to spend millions of dollars on the technology. But the business logic is sound and, over time, machine learning will be used to solve some of the industry’s tough challenges like the cost of compliance, the burden of paperwork and the need to combat fraud and money laundering.

Hedge funds and active managers are benefitting big time from machine learning. A variety of new machine learning-based companies promise investors new ways to identify trading opportunities, to sort through the noise to find the right data, and stay one step ahead of the competition by receiving trade signals earlier than others. Examples include Kensho, Dataminr, AlphaSense, Orbital Insights, and Lucena Research. Some hedge funds have built machine-learning technologies in house. Conferences, industry surveys, and press releases all point to hedge funds rapidly adopting machine learning-based technology.

Credit scoring and evaluating the creditworthiness of a consumer is another area of financial services primed for machine learning disruption. Tech companies believe machine learning can improve the traditional way of credit scoring by using AI to analyze large amounts of data and to improve the way customers are evaluated. In addition, the technology can be used for consumers with little or no credit history — those that don’t fit the traditional credit bureau model.

The 2010s saw a wave of innovation that led to the embrace of the term “fintech.” Loans, investing, payments, currencies, real estate and more were disrupted with new online-focused, low-cost models. As we approach the 2020s, the next wave of innovation appears to be a decade of artificial intelligence upending the financial industry.

The speed at which different industry verticals will adopt this technology remains to be seen and marketers have overused the term to sell products. But the business logic is certainly hard to deny. And some day, your voice assistant might be able to figure out that elephants can’t swim across an ocean.