Anyone who might have doubts regarding the extent to which technology has changed banking only needs to compare the trading floor of the mid-1990s to the current day. The noise and spectacle of open outcry trading has almost entirely been replaced by electronic trading systems. The change has been profound, with hundreds of years of tradition swept aside in a few short years. Trading floors will never look or sound the same again.
In comparison, the average neighborhood bank branch might not look like it's changed too much. There might be fewer of them, but most deal with customers with issues that they can't seem to solve online. These branches have a different function than they did years ago, and a tidal wave of technology is coming to continually revolutionize their function and role with the customer.
Artificial intelligence now has the potential to fundamentally change customers' relationships with banks, impacting every aspect of the life cycle from onboarding to retention and upselling.
Sure, AI is nothing new. And if it was going to change banking, wouldn't it have done so by now? Billions have already been invested in AI by banks and fintechs around the world, and many have yet to see a return. For smaller banks, the concept of AI probably seems like a pipe dream — what benefits will it really bring to their customers?
The problem is that historically, most AI has been very brittle. What we mean by this is that almost all implementations of AI have been built for a particular task, and that means the more places you want to use AI, the more "models" you need to create. It quickly becomes difficult and expensive to manage.
But this has changed. We're at the point where one model can complete a multitude of different tasks. It's a real tipping point for AI, and it's going to change banking too.
It's called a "foundation model" — a model that can be adapted to a wide range of tasks including document summarization, document classification, named entity extraction and sentiment analysis. But what's revolutionary is that these models are good — really good. To the point that it looks and feels like you're looking at the work of a human; it has levels of comprehension that have never been seen before, and it can effectively undertake tasks it wasn't trained to do.
Banks generate huge volumes of emails, phone calls, chats — all sorts of documents that tend to sit in digital storage. This is called unstructured data — it doesn't sit in a database and is difficult to analyze through traditional methods, yet rich with untapped insights.
Of course, it gets used — emails are read and replied to, phone calls picked up and handled, documents are read and distributed, but the insights at an aggregate level are hidden. If someone listened to the last 1,000 phone calls taken by a call center, what would be the biggest frustration they heard from customers? Having that insight could swiftly improve business processes.
When banks are limited to simply analyzing transaction data, it's sometimes difficult to read the tea leaves. But if you add the context of conversations and emails related to those transactions — not just what was said but how it was said, for example — we have insights that have never been captured. The potential impact is huge for an industry that runs on fractions of a percent.
Aggregate insight from unstructured data isn't just useful for understanding customer complaints. It means you gather next-level insights on risk, potential fraud, customer intention to churn, upsell possibility and trading opportunities.
When I ask banking leaders, "Do you use AI in your organization?" most of them tell me, "Sure we do." But when I ask them about their investment and drill into the details of what they're doing with AI, I hear a lot of frustrations. And I understand why — for years the promises have not lived up to the hype.
Sometimes it takes a step change for technology to make its mark. AI has fundamentally moved forward in a matter of months, and by the end of 2023, every bank will be figuring out how to leverage the latest developments. We have indeed reached that tipping point. Now is the time for banks to get smarter.
