The CFO’s Guide to AI and Machine Learning

The rise of artificial intelligence (AI) and machine learning (ML) has been a hot topic from board rooms to classrooms recently, dominating the discourse as people strive to understand its potential, opportunities, and risks. 

For businesses, AI and ML hold a powerful allure as companies consider how the new technologies might help boost productivity, cut costs, and gain a competitive edge. And for CFOs, AI will likely play an outsized role in helping the finance team continue its revolution to becoming more strategic, forward looking, and data driven.

In some ways, finance chiefs will treat AI like any new technology: balancing the need to be innovative with a healthy dose of skepticism around the technology’s limits, ROI, and inherent risks. Yet AI looks likely to drastically alter the playing field. As CFOs scramble to grasp the truly wide-ranging implications of AI technologies, it can prove difficult to cut through the hype and understand the technology as it stands today and the changes it might bring. 

In this special edition of the CFOLC newsletter, we’ll break down a few common AI misconceptions, explore possible AI and ML usage in finance, and offer considerations to keep in mind—giving you a base of practical knowledge as you move your business forward. 

Understand the Difference Between AI and Machine Learning

AI and ML are often used interchangeably, but machine learning is a subset of AI. An ML system uses pattern matching, models, and probabilities that are refined as it sees more samples of very specific data. 

For instance, a radiologist might use ML to evaluate X-rays. Does the system understand cancer research? No. But it can know, statistically speaking, what lung image aberrations look like after being given numerous examples of healthy and abnormal lungs. The algorithms that do the learning will not change unless a human alters them.  

Artificial intelligence, on the other hand is a more general term that applies to ML and other self-improving algorithms. AI seeks to make it possible for machines to behave in ways that both mimic and go beyond human capabilities. AI-enabled programs can analyze and contextualize huge amounts of data to provide information or even automatically trigger actions without human involvement.

All ML is a part of AI but not all AI is ML.

Where confusion commonly occurs is when AI is used to refer to every instance of smart algorithms. Classically, an algorithm is a set of rules that is executed when—and only when—it encounters a designated trigger. Every time that trigger is encountered, the algorithm will produce the same result. AI is a group of algorithms that can self-improve in response to ingesting more and more data. ML is more narrowly defined as analyzing only a single very specific type of data—chest x-rays and spell checkers are good examples.

Using Process-Driven Automation as an AI Foundation

As AI innovations proliferate, CFOs are looking toward implementing them to benefit the finance team. However, there are various levels of intelligent automation, and it’s important to start at the ground floor. 

Robotic process automation (RPA) is an example of a “ground floor” technology embedded in many software systems such as ERP. This business process automation technology can recognize, interpret, and classify business documents, storing their contents as data. For instance, accounts receivable and payable automation systems will use RPA to digitalize and classify paper and digital invoices.

RPA is not AI. RPA is considered “process-driven automation,” which uses business rules to accomplish tasks such as capturing and storing receipt images, classifying expenses, and enforcing reimbursement rules. This is different from data-driven automation, which is a more advanced level of intelligent automation depending on ML and AI to offer insights and guide decisions. 

As tempting as it may be to skip straight to the advanced levels of automation, you can’t skip the process automation phase. Before AI takes over, you will want business processes to be codified through RPA with humans helping along the way first. Once that is done, AI can follow the rules your company set to do the process with less help from people.

Focus first on identifying and automating the repetitive tasks in finance, such as rote tasks involved in matching of expenses, running payroll, and closing the books. Not only will automating those areas increase the efficiency of the finance team, it will also build a foundation for working with AI since the technology depends on accessible digital data and set processes.

Building AI and ML Into Finance

Once a finance team has incorporated process-driven automation, it makes sense to start considering opportunities to use data-driven ML and AI to further assist with and potentially automate operational decisions. 

A reality check is in order here: Many AI-augmented systems tend to be very good at handling just one thing—like offering customers smart upsells, or helping sales teams manage their clients. Going challenge by challenge to apply AI could result in a lot of standalone systems that limit your ability down the road to solve broader business management puzzles. 

Yet there’s a lot to be said for taking obvious quick wins, particularly if your business is slowed by one intractable problem. Steer toward an architecture strategy that creates a single data repository that can be used for one-off problem solving and lets you add new AI analysis and assistance as they emerge. Such a system should be capable of supporting complicated “what-if” data analysis and offer techniques such as data visualization and other helpful tools directly in applications. 

Combining finance data with operational data, web analytics, lead-generation data, store or warehouse traffic information, customer satisfaction metrics, and other business insights lets

analysts uncover trends unique to the company that wouldn’t otherwise be discoverable. And the right specialty system should be able to use that data store to accomplish its purpose easily—without a lot of custom code or extract, transform, and load (ETL) work. 

Want to learn more about AI and machine learning for the finance function? Download the “The CFO’s Guide to AI and Machine Learning” here.