Mention artificial intelligence (AI) to a person on the street, and you’ll conjure up Hollywood visions ranging from the humanity-crushing Skynet of the Terminator series to the robot love interests in Her or Ex Machina. Perspectives on the tangible impact of AI similarly range from Elon Musk’s declaration of AI as “the greatest risk we face as a civilization”1 to IBM Chief Science Officer Dr. Guruduth Banavar’s belief that “we’ve never known a technology that can have a greater benefit to all of society than artificial intelligence.”2 The reality, especially in the short term, likely lies somewhere between these extremes.
The quest for AI is not new.
Since Alan Turing’s seminal 1950s work on machine intelligence, AI has passed through several hype cycles. In 2020, we appear to have passed the crest of the current AI hype cycle, with AI enabling most major B2C websites and streaming services. By 2025, the market is expected to reach nearly $400 billion.3 This year, AI has been at the forefront of the coronavirus disease (COVID-19) response in California and China, tracking “shelter in place” compliance through cell phone data and potential viral outbreaks through air traffic and ticket data.4 The confluence of enormous computing power, globally distributed networks, and vast amounts of data have driven both the interest and potential of AI to unparalleled heights.
While most researchers believe that the pop culture version of strong AI—a sentient, thinking, human-like intelligence—remains decades away, “weak” or “narrow” AI is already here. Narrow AI, defined as non-sentient technology generally focused on a single task, has found its way into recommendation engines developed by companies like Amazon and Netflix, as well as endeavors as varied as insurance underwriting, automated call center response, and resume processing.
What if every
their own AI?
Is AI just for the big banks?
As with other major industries, financial services has been an early adopter of narrow AI, but the same cannot be said for the core finance department across industries. In a recent global survey, less than 25% of executives in the finance and accounting area felt that AI would enable significant staff reductions in the next five years, significantly lagging behind their peers in supply chain, HR, and even product/service development.5
Part of the barrier may be the perception that AI is “just for big banks.” Projects to automate fraud detection or identify money laundering across global financial networks take multiple years and deep pockets, so the common wisdom says that an organization needs the scale and resources of a global financial institution like Bank of America or the data volumes of Visa in order to reap ROI from an investment in AI. The finance departments of multibilliondollar organizations, let alone a finance department processing under $1 billion in payables and receivables, may not feel they can justify pursuing AI-powered projects.
As is often the case, the common wisdom is wrong. Finance departments outside the big banks have several clear use cases where the use of AI-driven technology makes both business and financial sense—and can deliver tangible ROI. Let’s examine a few.
A New World of Financial Analysis
How finance leaders can tap
into analytics to make business
more efficient and predictable
AI for fraud detection
While the big banks get most of the attention today, they are by no means the only ones using AI for fraud detection. Every finance department makes a concerted effort to detect data entry errors and/or internally-driven fraud, whether it’s over- or double-billing of customers, over-billing by suppliers, or more sophisticated embezzlement carried out through the creation of fake suppliers or customers within the finance system. Increasingly, AI technologies can be used to detect each of these different scenarios.
Additionally, in partnership with their IT organization, some CFOs have begun leveraging AI technologies to monitor hackers’ attempts to obtain customer or corporate financial data or to create false suppliers or transfer accounts.6 In the wake of many high-profile data breaches over the past several years, this practice has become top of mind for CFOs as well as CIOs. Should such intrusions occur, AIpowered technologies can also be able used to minimize the impact of data leaks.
But AI can.
AI for BI
The most forward-thinking finance departments have already begun to use AI to develop value and insight from their finance and customer data. For most organizations, the sheer volume, variety, and velocity of customer data precludes the type of traditional analysis done with existing analytic or business intelligence (BI) tools. Even big data or discovery tools too often rely either on sampling techniques driven by the creativity of a handful of data scientists or the establishment of costly data infrastructures with dedicated processing power.
Avoiding those more costly measures can help CFOs drive immense value, particularly in the realm of predictive customer behavior. In the U.S, according to the IRS, bad debt accounts for 0.5% of firms’ revenues. In 2018, that amounted to $100 billion in missing money, reducing profit margins by as much as 5%.7
Now consider credit risk and collections in 2020. For even small organizations, multiple years of customer and financial data can quickly add up to several terabytes of data. At this volume, humandriven analysis simply can’t operate quickly enough. But AI can. Using AI and machine learning to analyze patterns of customer behaviors that have led to late payments or bankrupt customers can yield immediate improvements related to collections and days sales outstanding (DSO). Imagine the value of not selling an expensive product or service to an organization that can’t or won’t pay the bill. For a small or medium-sized business dependent on short-term cash flow, choosing the right customers—and avoiding the wrong ones— can make the difference between success and failure, especially in times of uncertainty like the COVID-19 crisis.
Avoiding unprofitable customers is just one side of the financial equation: what about making sure you keep your best customers? In the new services economy, unsatisfied customers can quickly take their business—and your profits—elsewhere. Identifying the subtle customer behaviors that precede a profitable customer leaving can be an impossible task, especially when your financial and customer data lie in multiple, disconnected systems. Even if all your data resides in a Hadoop data lake, it could take a data scientist six months to formulate a subset of patterns that lead to customer churn. With the data volumes generated by modern CRM systems and the velocity of customer choice, AI will be critical for organizations seeking to identify, serve, and retain their most profitable customers.8
How the CFO can implement AI today
Organizations have several paths they can follow to bring AI into the finance department. As AI-driven analysis evolves from an interesting project to essential practice, it may be tempting to take the least risky option by building systems in-house. The challenge with “build-it-yourself AI” is that the user interface (UI) is the only simple part. Behind that UI, truly robust AI systems require heaps of intellectual property, large teams of data scientists and programmers, a massive amount of training data, labor-extensive data extraction and preparation, and last, but not least, an expensive technology infrastructure.
It’s not simply a matter of recruiting a few programmers from UC Berkeley or MIT to “build us some AI”. Even if it only took a couple smart developers, the chances of finding and hiring them would be slim for most companies. The world’s largest and most innovative businesses—e.g. Google, Amazon, IBM, Baidu, Salesforce, Facebook—are locked in an arms race for the best and brightest in AI. Meanwhile, those same companies are investing in and acquiring emerging AI vendors at a dizzying pace.
Some businesses may find value in working with some of the big names mentioned above, but it could be more worthwhile to find a vendor already using AI to solve problems specific to the practice of finance. Even independent AI vendors like Clarifai consistently beat the macro vendors around specific AI challenges like machine learning-driven image recognition for healthcare, video analysis, and ecommerce.9 As another example, collectAI combines your customer and financial data with their AI technology to optimize collections.10
The ideal solution for most CFOs, however, will be to simply pick a financial management solution with built-in AI. When onpremise systems were common, it would have been inconceivable to have AI built directly into the system. But expectations have changed now that cloud is standard. With cloud-based financial applications, the finance department should expect to see AI capabilities becoming more and more sophisticated in the applications they use on a daily basis. If your current finance vendor doesn’t provide a roadmap for incorporating AI into their system, it’s probably time to look for a new vendor.
Sooner than later, built-in, robust AI will become table stakes for the leading application cloud vendors. The challenge for customers will not be picking vendors or adopting the best AI technologies; the challenge will be reimagining how they do business today and in the future, how AI will change existing operating models, and how to best take advantage of the transformative capabilities of AI.
7. NYU Stern School of Business, 2019
8. https://www.salesforce.com/form/industries/financial-services/retail-banking-customer-loyalty-whitepaper.jsp 9. https://www.forbes.com/sites/aarontilley/2017/07/13/clarifai-ai-image-recognition/#2b16ca83fe42