If you’re not marketing that you use Artificial Intelligence (AI), you’re probably worried that you should be. AI is a go-to buzzword in business today. But I believe it is also one of the most misunderstood – both technically and from an application point of view. I feel many companies that say they are using AI are not. But I also believe that finance companies in particular have the opportunity to benefit from AI technology.

With this thesis in mind, I asked Professor Manjeet Rege, Chair of Software and Data Science at the University of St. Thomas, about the emergence of AI as a modern technology and its practical application in both industry-at-large and the finance ecosystem specifically.

Nelson: Professor Rege, how did you come to be an expert in artificial intelligence and the Chair of Software and Data Science at St Thomas?

Rege: My undergrad work was in mathematics, but mathematics alone did not provide an easy living at the time. I wasn’t interested in straight IT and saw a gap in how information systems worked with data so I pursued a PhD in Computer Science. Databases led me to data mining, a type of data analysis that today we call machine learning.  I completed my machine learning PhD in 2007, before it was cool.

There weren’t data science or ML positions yet which put me on an academic path. I was attracted to St. Thomas by the legacy and scope of the Graduate Programs in Software – Started in 1985, over 500 students with 300 in Data Science. St Thomas provides me a nice bridge between academics and practical application – my students are graduate students by night, professionals applying the technology by day.

Nelson: Interesting, from an occupational point of view you were a visionary with ML and AI – ahead of your time. When you put on your ‘data science historian hat”, was there something that “flipped the switch” on ML and AI after 2007?

Rege: Three important things triggered change in this timeframe:

1. Availability of data – raw material for ML and AI
2. Unrestricted compute available via the cloud
3. Fundamental advancement of learning algorithms and neural networks.

Due to these developments, there were a couple of small switches and then one big one. The smaller switches were successful examples of ecommerce – Amazon, Netflix, etc. I remember Netflix had $1M challenge to improve ML for their movie recommendation algorithms.

The big switch was around 2012 when Google did an experiment on YouTube videos. It was a unique experiment to switch from human tagging of videos to leveraging massive parallel computing automatically recognizing objects in the videos. Data scientists built models that had the ability to recognize patterns/objects – cats, dogs, and hundreds of other objects in the videos.  This demonstrated the power of deep learning/neural networks in a very practical and commercial application that showed other industries how ML and AI might work on their data – transactional, audio, etc.

Nelson: Your three triggers are now questions for individual companies to know whether they are ready for AI:

• Have we invested to capture our data?
• Do we understand how to use the massive scale of cloud compute?
• Do we have access to people who understand the necessary machine learning and AI tools like Azure Machine Learning and AWS SageMaker?

At St Thomas you have grad students representing local industries in retail, health care, medical devices, banking/finance, and of course manufacturing. Who do you see as most engaged with AI?

Rege: Everyone has woken up, but retail has been in the forefront.  Medical and healthcare have engaged mostly through medical imagery.  Using ML and AI on imagery has moved much faster than patient information applications. Financial companies have been the last to embrace AI.

We talked about how one definition of AI is that the software is allowed to experiment and learn from mistakes.  This is one reason it is challenging for financial companies – mistakes can be costly. Understandably, they see new technologies like AI as posing too much risk when dealing with personal and financial data. They have typically waited until they see others prove the “safety” of the technology before moving themselves.

Nelson: The ironic thing with your experience is that finance companies typically sit on tremendous amounts of data. The regulations and focus on privacy are like those in healthcare, yet you say they have lagged the adoption. Where have you seen adoption and what should finance companies be doing to leverage ML and AI?

Rege: I have seen good use of anomaly detection for fraud detection. Finance companies are essentially using the same pattern recognition models that YouTubes uses to accelerate and improve accuracy of these detection systems.

Second, we now have personalization capabilities, robo-advisors that provide Netflix-like recommendations for investors. There are two types of recommendation: product-based and user-based. The first recommends products are like the one purchased by their respective specifications – lawn mowers come with lawn mower blade recommendations. User-based recommendations learn from similar users and then recommend based on those experiences. The AI learns from other investors and portfolio, other consumers, and other mistakes. This is more of an optimization problem where the AI learns to achieve better outcomes.

Nelson: What about credit risk analysis?  The Wall Street Journal recently had an article talking about how big banks are moving away from FICO scores and using their own data.

Rege: Credit risk has also been improved with AI by moving away from just one number, e.g. FICO, because there are always many ways to arrive at that score. AI can leverage other parameters from that journey to look for anomalies and predict behavior to better inform the risk.

Nelson:  Does the St Thomas Business School ask you to help prepare its students for AI and do you find that your students enter the workplace better prepared than existing company leaders to leverage AI?

Rege: Yes, the St. Thomas Business School offers a master’s in Business Analytics for which about 45% of the courses come from the software and data science department.  The curriculum of business schools in general are becoming more technology centric each year.

To answer your second question, business leaders have the domain knowledge and students have technologic knowledge. This presents the question: Do I change or reskill to take advantage of AI? The correct answer is “And not Or.” Hire the data science talent, but domain knowledge from experience is the key to building any successful data platform. The corresponding need to reskill is one reason why the Graduate Programs in Software and Data Science have been so successful.

AI can no longer be a wishful aspiration for equipment finance companies. Each has the attributes critical to the commercial application of AI: large amounts of both historical and real-time data, availability of unlimited compute via the cloud, and access to services and ML and AI microservices.

Every successful application of technology begins by identifying an important problem to solve. Google needed to make searching videos easier and faster. Their solution unleashed AI into the broader marketplace. Ask what is limiting your growth and profit – operational efficiency, customer service, competitive differentiation, other? Once you have identified those problems, act knowing that you have the tools needed to solve them with AI.

About the Author

Scott Nelson is the Chief Digital Officer of Tamarack Technology. He has over 30 years of strategic technology development, deployment, and design thinking experience working with both entrepreneurs and Fortune 500 companies. Scott is a sought-after speaker and contributor on topics related to IoT and digital health. His involvement in technology in the local and national technology community reflects an ongoing and outstanding commitment to technology development and innovation.