How We are “Digitizing This Very Bank” at BNY Mellon

October 17, 2019

By Roman Regelman

The following is condensed from a keynote address delivered at the BankAI conference on October 7, 2019.

As organizations rush to embrace artificial intelligence (AI), are they approaching it the right way?

Banks traditionally see AI as this thing we put over here to fix some operational issue, address a specific use case or save money.

On the flip side, AI zealots think it will take over the world and require no humans.

They’re both wrong.

The future is not AI alone. It’s about unlocking the best of both robots and humans and using them in combination. We call that AI + HI: artificial intelligence plus human intelligence.

There are things computers can do better than humans – and vice versa. It’s not man against machine but both working together. When they do so, an organization achieves the most lasting, impactful performance improvement and creates disruptive innovation.

Within financial services, there are two main applications of artificial intelligence:

  1. Using AI to operate things better, faster and cheaper. This is the most common usage – but it’s just skimming the surface of what’s possible for companies like ours.

  2. Using AI to create completely different products and insights that are not possible without it. Most financial institutions are just scratching the surface of the potential uses of AI.

Early Adopters

At BNY Mellon, AI is not a sideshow; it’s embedded in our approach to digital. We are accelerating our uses of emerging technologies such as robotic processing automation, machine learning and artificial intelligence.

And we are not just experimenting with AI. We were one of the early adopters of robotic processing automation.

We have hundreds of robots in production that are performing repetitive tasks and removing the risks associated with manual entry. That allows for faster processing, gives us data that we can harness and frees people to focus on higher-value work.

Let me share a few examples:

  • Settlement fails forecasting service. BNY Mellon is in a unique position as the only bank that clears U.S. Treasuries. We are implementing a U.S. Treasury settlement fails forecasting service to ensure consistency and accuracy in treasury transactions. Using AI, we are looking at the data to see which transactions are not likely to settle. If we know that early enough, we can prompt clients to take action to avoid having the trades fail. With this information, humans can work with the clients to better manage their liquidity and make more efficient decisions.
     
  • Using natural language processing to read and direct emails. Our operations teams receive more than a million email inquiries a year, each of which they must read to determine the next best action and route the inquiry to the right team for resolution. That’s clerical activity that adds no value for the client.

We saw an opportunity to apply natural language processing to identify the intent of these emails so they can be addressed promptly. The machine learning solution determines what a particular inquiry is about, whether trade settlement, or cash, or a corporate action, or taxes, etc. It does that with 90% accuracy for the most common inquiries that represent more than half of all inquiries we receive all year.

To go even further, we are exploring cases where the machine can resolve a routine inquiry itself.

All that sounds like better, faster, cheaper. But there’s more to it than that. It’s allowing our people to spend more of their time doing what they’re hired to do – the work that is interesting, meaningful and moves the needle for our clients.

The application of machine learning in this case also gives us data that we can use to address root causes and reduce the need for those inquiries in the first place.

Let me take it one step further – we can also use AI to perform sentiment analysis. Humans have a couple limitations: they don’t have the ability to process thousands of interactions nearly as quickly as machines and their own biases act as filters. The sentiment data from AI can identify the areas that are frustrating to our clients, and our people can use that insight to take informed action. Together AI + HI can improve the client experience.

  • Digitizing contracts. We have more than 1 million contracts of any kind. The oldest contract is almost 100 years old, and it is an enforced contract. These contracts each have lots of content and each page has lots of unstructured data. We’ve now digitized them.

    Did AI make any lawyers disappear? No. But deploying AI to go through the documents to pull critical information has saved enormous costs and helped lawyers understand key aspects of the contracts, the consistency across contracts, and is helping to assure compliance and provide numerous client benefits. It’s artificial and human intelligence working together and allowing humans to create more value than they could alone.

Responsible Usage

As use of AI and data analytics in financial services has becomes more prevalent, it has set off a debate about how these technologies are used. As an organization, BNY Mellon is committed to ensuring that we always do the right thing for our clients, our regulators as well as keep our responsibilities to the wider financial system by virtue of the unique role we play.

We are working with the regulators on this front. Specifically, late last year, the Monetary Authority of Singapore published the Fairness, Ethics, Accountability and Transparency (F.E.A.T) principles to promote high standards within the use of Artificial Intelligence and Data Analytics in Financial Services.

We are joining a group of like-minded companies to validate the F.E.A.T principles systematically, as well as developing tools that institutions can use to validate their models against these principles in a standardized manner. We are specifically focused on anti-money laundering and countering financing of terrorism as well as regulatory reporting.

Second, we’ve thoughtfully constructed a diverse digital team – diversity is one of the lenses we apply in hiring. When you have a highly diverse team, you can expect that their awareness of bias is higher than that of a non-diverse team, which helps avoid unintentionally building bias into algorithms – in addition to all the other advantages of diversity in driving innovation.

 

Vast Possibilities

We’re just scratching the surface of the potential uses of AI.

By virtue of where BNY Mellon is in the industry, we have tons of data. AI allows us to sit on top of that and harness insights to not only service clients better but also turn those insights into decision-making tools for our clients and the global industry.

The possibilities – for us and our clients – are amazing.

 

Roman Regelman

Senior Executive Vice President, Head of Digital

For media inquiries, please send an email to Media Inquiries or view our Media Resources.