Will AI Revolutionize Regtech? Part 3: If You Build It
By Bo Howell and Cal Al-Dhubaib
Realize the value of building AI-driven fintech and regtech solutions and get practical advice for starting your own AI project.
The NSCP National Conference in November will feature a session titled “Will AI Revolutionize the Financial Services Industry?” This is the third in a series of related blog posts on AI and its impact on the industry. Part 1 covered the current state of AI development in fintech, and Part 2 highlighted financial services firms currently leveraging AI in their daily operations. Now let’s consider practical steps your firm can take to start an AI project.
Why Should You Care about AI?
AI is rapidly transforming our lives—from hyper-personalization to chatbots to quality control in call centers and beyond. Whether you’re already building an AI product or using software that has AI embedded in it, AI is no longer simply nice to have; it's an expectation. It’s easy to get caught up in the hype, however. So let’s start with a concrete definition: AI is software that can recognize and react to complex patterns to perform human tasks—like seeing, reading, reasoning, and creating.
AI is often viewed through the lens of automation. In reality, the majority of successful applications today are focused on augmentation. This means automating enough of the information sourcing in a task to allow a human to complete the more complex aspects efficiently and effectively. Here are five ways AI can give financial services firms an even greater edge.
In the past, designing your own AI-powered solution was exclusively the domain of big tech and required massive amounts of data. Three major trends are now changing the game. First, digital adoption has accelerated. Consumers initially hesitant to shop or bank online have rapidly shifted to digital interactions. This acceleration, in turn, changed human behavior and disrupted many preexisting predictive models, leading to a second trend: many companies are rushing to adopt more intelligent predictive models. Third, AI building blocks have become increasingly accessible. These building blocks include pretrained models, curated data sets, and tool sets that make working with data sources like text, documents, images, audio, and video more accessible than ever.
Building the Business Case
Regardless of your plans to develop your own AI-powered solutions, the first step you can take in this journey is developing AI literacy. Once you understand where AI can add value, you’ll be able to evaluate the capabilities of the many preexisting AI-powered software platforms. If a solution doesn’t already exist and it’s related to a core competency of your business, it may make sense to design your own.
Before starting an AI project, ask yourself these questions.
- What questions, behavior, or problems are you addressing?
- Who will be the product users?
- Who are the other product stakeholders?
- What data do you have to train the AI model?
- What types of algorithms should you use in your data?
- What metrics do you hope to achieve with your product?
All these elements go into your business case, which you’ll need to convince management, investors, and others to buy into your project. To start, identify your intent. What are you trying to address? Who’s impacted? Can you establish performance metrics to gauge success? When you can articulate the problem you’re trying to solve and whom your solution will help, you can then evaluate the available data and resources—including the capital, people, and other resources—needed to complete the project.
Once these pieces of information are gathered, you can start the product development process, which entails preprocessing the data appropriately, choosing the right models from many available options, tuning parameters in the model, and adapting the model architecture to suit the requirements of the application.
It’s a good idea to incorporate elements of product design thinking early into your project. MIT xPRO identifies four stages of the AI product design process: intelligence, business process, technology, and creation.
So Where Do You Start?
Many times, companies pursue “AI for the sake of AI.” It’s important to think about AI in terms of human enablement, not automation. This starts with understanding the task you’re intending to support. How is the task accomplished today, and by whom? Describing the task makes it easier to develop requirements for your use case. Here are some questions you can ask to get a better description of the task at hand.
- What steps would you outline when describing the task to someone else?
- What tasks are time consuming or particularly error prone? This question will help you identify opportunities to create value. Dig deeper into the why.
- Is the task time consuming because it’s repetitive or because it involves a lot of collaboration and information exchange?
- Is the task error prone because of complexity, lack of information, human fatigue, or general uncertainty?
- Is the task consistent? What you’re looking for here is whether subject matter experts generally agree.
- Is there missing information—or information that frequently needs to be found and that would make the task easier? This can help you identify opportunities for AI to provide insights.
- How often does the task repeat? And what value do decisions yield?
- How do you capture insights around this task today? What aspects of the task are represented digitally and which actions are tracked?
AI translators are professionals who are able to think about effective AI applications in the context of the above questions and identify situations where AI can add value. Follow the advice here to determine where precisely AI can add value to your own products and services. You’ll then be well on your way toward implementing effective—and even revolutionary—AI-powered solutions.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.