When it comes to tackling the increasingly complex problems of 21st-century banking like balancing digital transformation with regulatory compliance—all in a crises-ridden geopolitical climate—who better to partner with than someone who’s been in your shoes?
Before our VP & Head of AI, Paul Tepper, joined WorkFusion, he was Executive Director of Natural Language Processing at Morgan Stanley. We sat down with Paul to talk about his former role and how he leverages his banking industry intel today at WorkFusion to benefit our customers. This interview has been edited for length and clarity.
WorkFusion: To start, can you tell us about your past experience in finance?
Paul Tepper: I had been working for another vendor (Nuance Communications), selling software to banking and financial services customers. And then an opportunity came along to actually work at a bank. I was recruited by Morgan Stanley. I joined as an Executive Director on the wealth management side, where I led a team building conversational AI and search tools—largely internal tools. They had an internal contact center that had to field millions of calls, and we built tools to help automate those calls.
WF: At Morgan Stanley, was there any kind of direct relationship between your role and some of the roles of our customers? Did you experience the same kind of pain points as our customers?
PT: At WorkFusion, we use AI to automate repetitive knowledge work: anti–money laundering, KYC, sanctions screening, document processing, etc. One might not think of working in a contact center as knowledge work, but when someone’s calling and asking you a question—it is knowledge work. You have to answer questions, have subject matter expertise in that area to be able to answer the question, know how to look it up and find the answer, know who to contact to get the answer, etc.
In the contact center space, you’re looking at question-answering: trying to automate understanding the intent of a question, what are the specific entities inside that question (this is known as Natural Language Understanding or NLU). And at WorkFusion, we’re often looking at document classification and extraction — how to classify a document, what bucket to put it in, what specific fields to pull out of a document. The underlying NLP and ML technology is really the same. But to actually make it useful, it needs to capture the subject matter expertise of a real person, and a contact center agent and L1 analyst are very different jobs. The purpose of our Digital Workers is to do the job of a person—a specific person doing a specific role, not just any person.
WF: What motivated you to join WorkFusion?
PT: It was a combination of comfort zone in terms of my having dealt with enterprise sales as a vendor, having worked in banking and financial services, but at the same time it was a good learning opportunity for me. I hadn’t worked in this specific kind of automation space (AML, KYC, sanctions screening) or in document processing before. It was the right balance of familiarity so I could hit the ground running and understand what we needed to do and why, while at the same time, the right number of areas to grow and things to learn.
WF: Why and how are you helping to solve problems you used to face in your former role?
PT: I spend a lot of time dealing with model risk management, or MRM. This has to do with the risk financial institutions take on to their business in deploying machine learning models to make decisions for them and to automate processes. That could be anything from deciding what trades or investments to make, to whether to approve somebody for a loan or a credit card—those really high-impact decisions—down to less risky decisions, like whether to show somebody this article or that article when they’re searching for the answer to a question.
MRM is all about cataloging and governing these models, and assessing the risk each of them poses to the organization. That’s directly the kind of things I did in my previous roles and I spend a significant amount of time helping our customers with MRM at WorkFusion: developing clear documentation for our users around these issues, helping to build models that are explainable so they can satisfy their requirements, and actually having these conversations with our customers so they can get their requirements met and ultimately deploy the models. So I would say my former roles been very helpful.
WF: How do our customers benefit from the fact that you used to work at a bank?
PT: The number one thing is my MRM experience. I’m often brought in by our Sales and Account teams to help deal with that issue, to help guide them through what can be a difficult and onerous task. It’s difficult because it’s really important. You have to cross all your T’s and dot all your I’s when you’re looking at these critical issues around bias, explainability, and risk in your models, which can be making very important decisions, e.g., whether or not someone is flagged for being on a sanctions list. That’s probably the number one area where having done that on the other side myself, being the one who had to interview the vendor and try to gather the information needed to satisfy our risk requirements and privacy, is now a benefit to our customers. I was on the other side trying to look at several different vendors to solve a specific problem and evaluating them. So I can see it from both sides. I can empathize and say, “OK, I get that this is the information that you need and we’ll try to help you get the information so you can make the best decision.”
Then there’s deployment issues. There are a lot of processes that have to be in place to make sure that the deployment isn’t going to be a risk to the business. For example, when we release a new feature to a customer, we need to know if they are going to be able to accept it? There’s a variety of ways that my experience of living and working in their shoes has helped me empathize and helped the customers that we work with, which you can’t really know unless you’ve done it yourself.
WF: What do you think is the biggest problem that WorkFusion is solving for banking today and looking ahead a few years?
PT: There’s a macro and a micro answer. At a macro level, banks are spending a lot of money on technology. The big ones have tech budgets in the billions of dollars, so it makes them sometimes feel like they’re tech companies. But when it comes down to it, they’re not tech companies. Their primary business is banking—taking care of people’s money, being good stewards of their capital, making good investments. They have compliance issues to deal with. So they can’t operate like tech companies. They can’t move as fast as a tech company, and that’s by design.
We are a tech company that empathizes with them and speaks to them by offering out-of-the-box solutions so they don’t have to start from scratch and build their own internal data science or AI team. At the same time, we’re building these products in such a way that you can still customize them for your business.
The micro level is the sanctions screening space. We’re making a big dent in the deluge of alerts customers are dealing with in the sanctions screening and anti–money laundering space. Right now, it’s overwhelming. You may say, well, how does AI impact jobs? That isn’t really the problem they’re dealing with. They’re dealing with the fact that there’s so much information, so many alerts coming in that things can potentially slip through the cracks—and that’s really problematic.
Looking ahead, our strategy is to form kind of a network of Digital Workers across customer banks where they’re deployed. If customers choose to opt into this network, they can benefit from the network effects—completely securely and privately. No data will ever leave your particular installation, but you’ll get the advantage of not working alone on a solution.
To hear from more WorkFusioners who used to work in banking, read previous installments of our Q&A series “From Finance to Fintech”:
To learn more about our AI-enabled Digital Workers, please schedule a demo.