TechTalk: The Evolution of Artificial Intelligence in Financial Technology

How It Works, What It Does, Where It's Going

With John Purcell, Esker Senior Value Engineer

In this first TechTalk article, we explore artificial intelligence and its use in financial payment software. While everyone has probably heard of AI and understands it’s being used in just about everything, we wanted to understand how it works specifically within AP software.

We spoke to subject matter expert John Purcell, Senior Value Engineer at Esker, to get his take on how this technology is being used in his organization.

John, let’s get started by discussing how you got started in financial technology. What was your first role and how has that evolved over time? What are your current responsibilities at Esker?

My career in fintech began 18 years ago when I took on a role as a project manager at an enterprise pricing and revenue management software company. In that position, I not only managed customer implementations, but also delved into the forecasting, segmentation and optimization techniques the software used to suggest prices. It really sparked my interest in advanced analytics, and I transitioned into consulting after a few years. In those roles, our teams developed tools and value cases for Fortune 100 companies, on topics ranging from supply chain to pricing. Those experiences honed my expertise and introduced me to the office of the CFO focus I now have at Esker.

Here at Esker, I’m the Senior Value Engineer specializing in our source-to-pay solutions. My role involves supporting our sales team through business case development, ROI rationalization and all aspects of value creation. By aligning our solutions with client needs, I help them understand and quantify the efficiency and cost savings achievable through Esker.

What was your first exposure to AI and RPA, and what was your initial impression?

My experience with RPA dates to my first job in 2001, where I built macros to automate various tasks. It freed up so much time for me to work on more interesting, strategic initiatives, and it also made me look pretty good to my boss. More recently, I have leveraged screen-scraping RPA technologies to gather market information for various analyses.

Screen scraping is an automated way of copying information from a website by programming a computer to navigate to the website, copy the data and then store it for you. I specifically benefited from it when we wrote a program to grab the prices of a group of automotive parts from various websites daily and feed that information into a pricing model.  

As far as AI, my first exposure to that was as a consultant in 2013, where I worked with several operations research experts holding master's degrees and PhDs. We applied machine learning to forecasting models, identifying patterns that significantly improved our results. Our customers were truly amazed by the technology’s impact on their forecasts. As our lead data scientist liked to remind us, “Machine learning is nothing new; it’s the same math from 100 years ago, but computer processing power has made it a reality.” Twelve years later, the capabilities continue to amaze us and innovations allow us to harness the possibilities.

These experiences have shown me the immense potential of automation in enhancing efficiency and accuracy.

Can we take a minute to explore what machine learning is, exactly?

I think of machine learning like teaching a computer to learn from its “experience,” just like you or I do. You can do this by feeding the computer loads of data and examples so it can figure out patterns and make its own decisions. Take photo recognition for example. If we wanted a computer to recognize a dog, we would supply it with millions of photos of dogs and other animals. Over time, the computer learns the features that make a dog a dog, and it gets better at identifying a dog in a picture that also contains a cat.

Let’s talk for a minute about the differences between AI and RPA and how they work together. What should a non-technical user know about them?

Sure, it’s not always clear where one starts and the other ends. RPA is ideal for tasks with clear, structured rules. It’s programmed to mimic human actions that perform well-defined business functions. I think of it as a digital worker that follows a set of instructions to complete routine tasks like data entry.

AI handles the unstructured data and complex processes that RPA can’t. It includes technologies like machine learning and neural networks that are designed to act like a human brain. It's like a smart assistant that understands and adapts to data inputs, like when your smart vacuum cleans your living room or the Waymo Driver takes you to your destination.

When you combine them, it makes it possible to automate an elaborate process like the cash conversion cycle. RPA efficiently handles the repetitive tasks, while AI enhances RPA by managing the scenarios and unstructured data. Imagine processing invoices: AI can recognize and process both standard invoices and those with an unusual format, then an RPA bot will create the invoice in the ERP and trigger payment once the approval workflow is complete. Together, they create a powerful automation solution, leading to greater efficiency and accuracy.

Did you realize early on that AI and RPA would be a game-changer for Esker? What was the process like for incorporating these advanced technologies into your solution?

Esker recognized early on that AI and RPA would be game-changers for our solutions. As early adopters, we have been building and refining our own AI models for many years. This foresight allowed us to seamlessly integrate AI into our solutions. Our AI is built, trained and maintained to handle tasks specific to the cash conversion cycle, ensuring the efficiency and accuracy that is so valuable in our space. Building these capabilities into our solutions involved extensive research and development, continuous training of models, and rigorous testing to ensure seamless implementation and function. As a result, our solutions are not only advanced, but also highly reliable, which is a key part of our value proposition of streamlining operations and enhancing productivity.

What benefits do these technologies offer to users? How can people best leverage them for their own organizations?

Individual users will benefit from escaping the mundane to focus on more meaningful and gratifying work. Personally, I’ve found tools like ChatGPT and Copilot to be a huge help when getting started on an idea or project — they’re my personal sounding board!  Users that invest the time to learn to be experts in prompting the AI tools will get so much value out of it, and those that can build their own tools on top of that technology will superpower their work output.

Going a layer deeper to software specific AI solutions, the insights and automation available in solutions like Esker’s are immense. You’re able to do things like removing hours of analysis and prep time for that monthly report, converting your paper documents to electronic as they are received with little or no oversight, and querying your systems in real time without needing an advanced degree. To that end, it will be important to have AI specialists and frameworks in place to take advantage of the many possibilities.

Where do you see the future of AI and RPA going? What further evolution do you envision?

Adoption of AI and RPA will only increase as these tools become more advanced and, more importantly, more intuitive to use. For instance, user-friendly interfaces and natural language processing capabilities are making these technologies accessible to a broader audience. Here at Esker, we prioritize research and development, and are continuously looking for ways to improve our products and enhance our customers' experiences. Currently, AI is the tool we believe will have the most significant impact. With advancements in generative AI, such as GPT-4 and the like, we are seeing more sophisticated applications that can understand and generate human-like text. It's only a matter of time before we see agentic systems — AI systems capable of autonomous decision-making — becoming commonplace.

However, let me share a story that I think will give some perspective on where we are with AI right now.

Last summer, I used ChatGPT, an AI writing tool, to create spooky campfire stories personalized for my children on a camping trip. The initial draft was built on the basic information I gave it — my kids' names, the setting for the story and that I wanted it to be in the spirit of classic campfire stories. We didn’t think the first draft was all that scary, so I asked to make it scarier. Unfortunately, the second draft wasn’t much different. It simply added adjectives and adverbs to the story so that “It was a dark and stormy night” became “It was a very dark and stormy night.” Ultimately, I learned to give ChatGPT specific instructions and context to craft the story we wanted. I think that experience epitomizes how we will use AI to enhance our work and careers, not replace them.   

Which calls into question whether financial practitioners need to be concerned about AI taking their jobs. I don’t see a need for worry, but I do see an opportunity to adapt and develop new skills with AI as a tool that can help us achieve our potential. That is how I approach it in my value engineering role at Esker. AI helps me research topics, validates calculations and assumptions, and proofreads my content in many cases, but it’s just a tool that I am learning to apply to my situation. As AI improves, I believe I will as well.