The two words on just about every business leader’s mind right now are artificial intelligence. Recent advances suggest that the cutting edge is only the beginning of what AI tools will offer—and though we’re still in early days, decision makers across industries are already looking for how they can use AI to solve business problems of all sizes.
But as the saying goes, when all you have is a hammer, everything looks like a nail. I’d say that it behooves just about all business leaders tasked with AI implementation to take their time in assessing effective use cases, so as to avoid inventing problems—or nails—for it to solve. We are still in the earliest stages of corporate AI usage and many likely developments have yet to be borne out. Overall, while there are many areas in which AI solutions show promise, there are some places where AI just doesn’t make sense (at least, not yet).
Many of the most effective uses for AI today are less dramatic and revolutionary than the current climate might suggest—but that doesn’t mean their benefits are less substantial. This is certainly true in the electronic bill payment and presentment (EBPP) space. AI can be highly beneficial when used to mitigate common or everyday practical EBPP pain points—particularly in B2B use cases. Though these applications of AI may seem subtle, when taken in aggregate, they could have a huge impact.
Customer Support
An easy place to start with AI is with generative AI, which is becoming more mainstream every day. Platforms such as Microsoft-backed ChatGPT and Google-backed Bard can “read” and “answer” questions or prompts written in plain language by analyzing the data they have been fed and producing responses based on information they deem relevant. In the cases of ChatGPT and Bard, that data comprises billions upon billions of websites and texts available online; in more proprietary use cases the AI can be taught on selected data.
Generative AI, particularly when Natural Language Understanding (NLU) is implemented, could prove to be invaluable in many customer-facing operations in electronic payments, especially when it comes to biller inquiries. EBPP is fairly straightforward on paper, but the tools can come across as unnecessarily complex for customers when front-end payments systems are modified by multiple integrations, as is often the case for B2B payment companies that serve many industries.
Training a generative AI with NLU on historical customer inquiries or roadblocks could allow it to instantly respond to common questions, like those about identifying specific charges. This gets customers the information they need without requiring them to wait for human assistance, saving the customer service team time and resources, and creating a positive and prompt customer experience. Not only that, the AI learns from each interaction, gaining data that can better equip it to analyze, adapt, and improve its responses to future customer support inquiries. And all of this can be done with caution around sensitive data at the fore.
What’s more, AI is available 24/7 and can be trained for use in many different languages. This can be helpful for EBPP companies with international customer bases.
Parsing Data
AI is an incredibly effective tool for automating and improving the accuracy of tedious, rote tasks—and there are few things more tedious and rote than invoice matching. Even worse, it’s a task that demands exact precision, and one where imprecision can have huge consequences. AI can accurately cross-reference countless minute details of invoices and payment data in the blink of an eye, flag anomalies, and quickly identify possible fraud or error.
And again, AI is continuously learning: When such anomalies are corrected, AI can digest that data to further refine its matching algorithms. Continuous learning also allows it to discern patterns in historical anomalies, and thus identify commonalities that could point to potential risk factors. Once these red flags are raised, companies can implement corrective measures.
Trend Analysis and Platform Resiliency
Like invoice matching, the logging and analyzing of data is another task that, when done manually, is tedious, time-consuming, and prone to error. An AI algorithm can not only automate data-logging, but can simultaneously analyze that data for anomalies or trends around things like transactions and customer behavior. This information is obviously invaluable to (human) EBPP decision makers who are developing corporate strategy or general forward-looking plans. But it’s also valuable to the AI, which can be trained to alert security teams when it identifies discrepancies that may indicate an incident.
Speaking of alerts, AI can also reduce what IT teams call “alert fatigue,” which occurs when an oversaturation of alerts winds up having the opposite of the intended effect. When alerts happen all the time, systems administrators can become desensitized to truly critical contingencies. AI can assist IT teams with overall platform resiliency by continuously monitoring server health, the network traffic, and transactions that affect it. It can also analyze historical data around server downtime to predict the conditions under which systems may be overloaded, allowing teams to proactively devise workarounds for those situations.
Fraud Detection, Prevention, and Overall Data Security
Among the most critical concerns across departments in the payments industry—or anywhere in the financial sector, for that matter—are fraud detection and data security. Everyone wants to feel confident that sensitive data like credit card numbers or banking information is secure and can’t be accessed by bad actors. This is another area where AI’s unparalleled pattern recognition can be invaluable. Having “learned” historical customer behaviors, for instance, AI can flag, in real time, anomalous activities around access patterns, which could indicate security incidents.
Identifying these occurrences in real time can allow security teams to act quickly and protect targeted data. By analyzing historical information, AI can spot potential fraud—be it suspicious transactions, unusual logins, or anything else—more quickly than even the fastest human. Even as bad actors react to increased security, devising novel tactics for circumventing defenses, AI will always be simultaneously improving and adapting to their behavior.
While the AI solutions above may not be the stuff of futuristic science fiction novels, when taken together they could make the electronic bill payment and presentment industry even faster and more reliable than it is today. Of course, incorporating AI in these ways doesn’t eliminate the need for human participation. Far from it: It’s vital that any AI incorporation involves careful strategy—the kind only a human can think up.
