Now there are people who would argue that blockchain technology or the distributive ledger is the most important new technology, and I certainly think that both of those are transformative. But when it comes to finding a really good fit between the customer and the banking industry, there is no more transformative technology than artificial intelligence.
Why Is Artificial Intelligence (AI) Transformative?
Take a look at the JPMorgan COIN technology. COIN, short for “contract intelligence”, is the reason why a February 2017 headline claimed “JPMorgan Software Does in Seconds What Took Lawyers 360,000 Hours” That means that if you assume the average work year is around 2,000 hours, the JPMorgan intelligence software replaced the work of 180 people.
This is the direction that technology is heading, and it is one reason why AI is likely to be a major game changer. JPMorgan didn’t just double or triple their throughput on these contracts, they exponentially improved their contract processing.
Their COIN software is a learning machine built to parce financial deals that would’ve kept a legal team busy for thousands of hours. The machine learned to read the words of the contracts. It learned to understand what the lawyers were looking for, and then was able to go through and evaluate all of these things without having to hire people.
What happened to those 180 people whose work is now being done by intelligence software? Unfortunately, some of them may have been laid off. And yet it’s also possible that those people were simply freed up to work on higher value tasks. In reality, it was probably a combination.
Bringing a Knife to a Gun Fight
In the example of the COIN software, AI improved process all the way around. It radically increased efficiency, effectiveness, and throughput. Not only does increased throughput equal increased revenue, the fact that COIN now does the job of 180 lawyers means that the company either saved money on hiring people, or was able to assign its employees to more productive work.
You can imagine how AI software like this will dramatically increase the margin between companies with AI technology and those without. A day is coming when not having AI software will be like bringing a knife to a gun fight.
In the history of war, the development of guns was an exponential differentiator. There was no turning back to older weapons and the focus became bigger and better guns. Artificial intelligence is going to become an arms race as time goes on. The problem is that many credit unions and other financial institutions are not even in the race.
What Makes AI So Important?
A lot of people think that the rise of AI is due to networking or processing power. Something like Moore’s law, a 1965 prediction that claimed the number of transistors per silicon chip would double every year.
But the real power of AI is data. Organizations have been dealing in data for a long time. Look at Amazon, I mean, I can’t even think about buying really important purchases any more without at least looking at reviews. That’s a use for data that Amazon has moved into our culture. However, even though they use data in unbelievable ways, in the end, all of their data is still mostly limited to how people interact with their company and their products.
Banks and financial institutions have a much broader data landscape. Think of it this way. Home Depot knows what you spent at Home Depot and Lowes knows what you spent at Lowes. The bank is the only one who knows how much you spent at both places. They also know how much you spent on your home, how much you spent on your car, and how much is in your billpay.
Essentially, your financial institution knows more about you than any other retailer on the planet. Facebook, Twitter, big data companies, they still come in second to banks. We know more.
The reason we know more stems from one simple fact:
You are what you buy.
Because people are what they buy, funneling this data into artificial intelligence software has the potential for a vast number of useful, customer-centric services.
Cats and the Vast Potential of Machine Learning
Customer-centric services brings us to the topic of cat pictures. Imagine trying to teach a computer to recognize a cat. One option would be for the programmer to go through and tell the computer to look for whiskers, four legs, fur, a tail—all those characteristics that make up a cat. And ultimately, the computer would be taught based off of a human interpretation of a cat.
But what if instead of that, we told the computer itself to learn what a cat is. What if we said, “Here’s information from Wikipedia. Here is data from 60,000 cat websites. And, of course, here are the 400 bazillion cat videos from YouTube. Now you decide what a cat is.”
In telling the computer itself to decide what attributes distinguish a cat, you are opening the door for the computer to comprehend and make distinctions on a level that no human brain could connect. Its vast processing capabilities will look at things that no one would conceive and find correlations that we could never put together.
This gets to the root of why AI has such useful potential. If you think about our data, we have insane amounts of information about people, including what they buy and when they buy it and how they buy it. The question is, how can we build a software that will use that data in a way that benefits the customer.
Customer-Centric Artificial Intelligence
As much as I’m interested in what JPMorgan did on the backend, and what they did internally, I’m much more interested in how we can make this something useful for the customer.
In that vein, what I’m seeing is the rise of the personal banking assistant. In this case the idea is to teach an AI by feeding it your bank account data so that it can understand your background and look for trends. Essentially, the idea is to equip it with data so that it can act as your financial assistant.
So for example, your AI assistant could analyze your credit card system and tell you that adjusting your payment by a certain amount would raise your FICO score by 30 points. Or maybe while you are out water skiing, it will poke around on the web and find that your car has a recall. Maybe you just purchased some items and your AI financial assistant lets you know that there is a cheaper price available so you can get a refund.
I call these examples of how AI could be used intuitions. Right now I don’t see these things acting without a lot of training—people are going to be pretty fearful of that. But over time they may be trained to an extent that they are trusted with some autonomy.
Maybe your AI notices that you haven’t made the HOA payment that you normally make at this time of the month and asks if you want it to take care of the payment for you. If you intended to make the payment and simply forgot, you might tell your AI to go ahead and take care of the HOA payment each month automatically. So you learn to trust it with certain tasks.
Another aspect of AI usefulness is scenario planning. The world is changing so quickly and determining where, when, and how to conduct business is increasingly complex and increasingly important. Feeding data into an AI gives it an opportunity to analyze scenarios with a processing power that goes far beyond a team of experts.
So if you were looking at what to study, if you were sitting around at your bank saying “Hey guys, what are we going to spend the rest of 2018 looking at?”, artificial intelligence is the answer. Buy the books, go to the conferences, read the blogs, whatever it takes to focus on AI. Because getting a head start on its transformative technology is going to prove very valuable.
Back links to referenced content:
“JPMorgan Software Does in Seconds What Took Lawyers 360,000 Hours”: https://www.bloomberg.com/news/articles/2017-02-28/jpmorgan-marshals-an-army-of-developers-to-automate-high-finance [Editor’s note: this article is restricted to Bloomberg professional service subscribers. I was not able to access the full article, so this may not be the best hyperlink. Here is one option if you want an article that everyone can access without a subscription: https://futurism.com/an-ai-completed-360000-hours-of-finance-work-in-just-seconds/ Obviously, if you link to a different article, the title and date in the post should be changed to match.]