I first learned about artificial intelligence (AI) sitting in a computer science class at University of California, Berkeley. I was immediately hooked.
Back then, though, it was more of an intriguing and futuristic concept for building machines capable of evidential reasoning and autonomous machine learning. It would take at least two more decades before computer processing power and machine learning approaches would catch up to the point where it could support practical AI business applications at scale.
More than 12 years ago, I arrived at Oversight and I saw how AI and robotic process automation (RPA) solutions could deliver significant business value by finding hidden financial risks and fraudulent spending that would have taken armies of auditors months to uncover.
Today, I am more excited than ever because enhancements in machine learning are accelerating the rate of AI’s adoption and with it, the ability to transform the way we detect, mitigate, and even prevent financial risks.
Transforming finance processes
In my view – and those of many industry experts – 2018 will be the year that AI finally gets to work in enterprises on a broader scale.
According to a recent survey by Grant Thornton, CFOs ranked advanced analytics, AI, and RPA among the top planned investments. 90% of CFOs believe AI will transform the finance function over the next five years.
Here at Oversight Systems, we are seeing growing interest in and adoption of our automated, AI-powered risk management and compliance monitoring solutions for procure-to-pay, travel and expense, and purchase card processes.
Two primary factors are driving the increasing adoption of AI in finance, risk management, procure to-pay, call centers, compliance, and IT:
Processing power. We have arrived at a point where we have the computing power and calculation speeds capable of crunching hundreds of simultaneous, complex analytics. In other words, you can write the most complex and powerful AI and machine learning algorithms. However, without a computer processor powerful enough to support complex, multi-level analysis and feedback loops, you will not obtain full value from the application.
The maturing of machine learning. Machine learning algorithms have come a long way, too. They have moved beyond the novice ability to pit a machine against the best human players of Jeopardy. Instead, we have evolved to the point where Google’s AI-powered AlphaZero mastered the entire game of chess in just four hours. And, in that time, it didn’t just master the rules – it figured out how to beat anyone or anything – without human knowledge or intervention. And Google’s AI, repurposed as AlphaGo has repeatedly beat the world’s best Go players.
Improvements in machine learning algorithms aren’t just revolutionizing the ancient games of chess and Go – they’re reinventing enterprise intelligence. Machine learning has matured to the point where targeted enterprise applications are practical, available, and reliable. In short, they are delivering real business value.
Take Oversight as an example. Starting back in 2003, we pioneered the use of patented AI, guided machine learning, and evidential reasoning technologies to power the analytics engine of our Insights On Demand® solution.
Uncovering hidden fraud and waste
Fifteen years later, many of the world’s Fortune 1000 companies and government organizations rely on Oversight’s automated auditing and compliance monitoring platform analyze every single financial transaction to detect hidden patterns of fraud, waste, and misuse. Over time, the quality of the analysis improves through both guided and unguided machine learning, so the system gets increasingly smarter.
While the algorithms used to understand the full context of transactions are complex, the Oversight platform delivers plain-language insights to users via a patented workbench. It does not require line of business managers or executives to become analytics experts. For example, AI can be used to:
- Detect external fraud by identifying suspicious invoices with an abnormal invoice number pattern, payment currency, bank account, and address change. The AI system uses evidential reasoning to analyze multiple dimensions of risk, prioritizing high-risk invoices for user review.
- Track and analyze meal and entertainment expenses over time across multiple employees to detect abnormal patterns such as frequent entertainment of an attendee such as a government official based on factors like excessive expense amounts for a given geographic location.
The blend of AI, mature machine learning, and a user-friendly workbench enables managers and executives to automate complex and repetitive business processes, savings hours each day while improving overall productivity. It is this combination that will drive more enterprises to deploy packaged AI finance applications in 2018 and beyond to better manage risks, improve operational efficiency, and seize new opportunities.
AI will become more useful in 2018 and beyond. Business leaders should engage now with providers that have a proven record in enterprise AI applications to help them deploy the most impactful solutions for machine-fueled insight and business process improvement.