As CFOs, we have a full plate of responsibilities for financial reporting, planning and analysis, and risk management. We’re often skeptical of “breakthroughs” that promise to “transform the finance function” and “shift the paradigm of accounting and finance.” To us, these types of phrases just sound like more of the same – another overhyped management consulting fad based on a Harvard Business Review article. All we hear is just “blah, blah, blah” and we tune it out.

Should we feel that way about the current wave of Artificial Intelligence as well?

The name itself suggests, AI is a technologically derived intelligence. But AI models mostly just repeat what they have been taught. They don’t really “know” things or think like people do. When truly original “thinking” is required – such as creating an original, testable scientific hypothesis – it turns out that current AI models are quite dumb.

Looking at it through this lens, is AI a bane or boon to today’s CFOs?

The Discussion around Artificial Intelligence

Forms of artificial Intelligence have been around for decades. Machine Learning (ML) uses models or algorithms that have been “trained” on large datasets to create human-like functions, such as quality control on manufacturing assembly lines. CFOs with credit card companies have used ML to train their models on past customer behavior, including default patterns, to set or modify credit limits. Another form of AI is Optical Character Recognition (OCR), which can capture data from documents and convert to machine readable text. OCR was perfected by IBM in the 1950s and is now used in a wide variety of business applications.

What’s new in the current wave of AI is the emergence of generative AI models, which can seemingly create new content. Using Large Language Models (LLMs), applications such as ChatGPT can “predict” and compile words and sentences, based on past usage.

What’s also new is that generative AI models can be trained on images, not just text. It is becoming increasingly difficult to identify truly original content that has not been generated by an AI algorithm. People often see an image and ask themselves, “Was that photoshopped?” As a result, there is an emerging market desire for a certification of original content. Manufacturers of photographic equipment have developed cameras that apply a forgery proof signature to photographs that certify the image’s originally produced content. Similarly, blockchain functionality could present a potential method of verifying/validating the chain of custody for financial documents.

How Can CFOs Leverage AI?

There are several use cases where AI can augment humans in the finance and accounting function. Here are a few common examples of those types of AI applications:

• Fraud Detection

ML could be used to develop algorithms that recognize unusual behaviors in selected general ledger accounts that could be used to detect fraud.

• Better Audit Testing

For certain audit procedures, testing involves a random, representative sample of contracts as an example. Sampling is often required because the contracts are not standardized, and contract terms must be entered manually. Using improved OCR, auditors can cost effectively take a 100 percent sample.

• Document Drafting

Generative AI can be used to prepare a first draft of a document by leveraging the existing documents in the marketplace. Depending on the document (for example, a public 10K filing, transfer pricing study, or business continuity plan), this process can capture industry trends, use the right industry nomenclature, consult a checklist, and cite key performance indicators that the CFO authoring it may not have considered.

The Risks of AI

CFOs have long managed the risks of ML, which mostly center on an incorrectly specified model or algorithm. However, generative AI presents new risks that must be managed by the CFO (who is often also the chief risk officer), including:

• Hallucinations – untethered text generation or synthesized ungrounded text.

ChatGPT makes up stuff. A linguistics professor created a nonsensical phrase “argumentative dipthongization” and presented it to ChatGPT, which proceeded to spew out four paragraphs of content, including an entirely fabricated mention of the person first credited with using the term.

• Infringement – authors of original content may have copyrighted their work.

Recent WSJ article 07/30/23 asked if content “generated” by ChatGPT could be considered a “derivative work” and therefore subject to charges of IP infringement.

• Material nonpublic info – relevant and important company information that has not yet been made public.

How can companies keep employees using ChatGPT from inadvertently sharing material, nonpublic information? This can happen when AI models are initially trained on internal nonpublic information and then used to produce publicly available documents.

• Advanced phishing scams – bad actors using generative AI to build more effective phishing attacks.

Fraudulent emails can more convincingly mimic the “voice” of the compromised employee to make them much harder to detect.

Should CFOs Embrace AI?

Yes, today’s CFOs should embrace AI. There are numerous AI tools that can improve the efficiency of the finance and accounting function, which will enable the CFO (and the entire finance and accounting team) to spend more time on predictive analysis and forecasting and less time on information inputting and processing.

According to ITR Economics, the current shortage of knowledge workers is not expected to be filled by growth in the US population or immigration. This labor gap can only be closed by improvements in worker productivity, which will be aided in large part by AI. However, the CFO must maintain human oversight of the finance and accounting functions. AI is a tool to augment the work ultimately done by humans. AI has no moral compass. AI has no ability to make judgements or do the right thing, even when it flies in the face of the “data.” Thankfully, this is still done by human beings.

Next Steps: When you need outsourced CFO services, we can help! Our highly experienced fractional CFOs offer executive-level financial leadership in a less-than full-time capacity, for interim, part-time, and project-based engagements. Contact us to find out more today!

About the Author
Larry Breitbarth – Partner, CFO Selections

Larry Breitbarth is an experienced financial executive with a proven track record in financial management, operational and strategic planning, performance improvement, corporate controllership, project management, and valuation consulting. He began his career as a business valuation consultant with Arthur Andersen in Chicago where he served manufacturing and distribution clients in the automotive and food equipment sectors.

Prior to joining CFO Selections, Larry held senior financial positions, from Senior Manager to CFO, with a variety of companies in the technology and financial services industries. Larry has also been a successful consultant. As a senior manager for Big 4 Consulting firms, he led valuation consulting projects (including projects for Nordstrom, Starbucks, and Microsoft), testified as an expert witness in valuation disputes, developed and managed remote regional offices, and generated over $50M in tax savings for clients through the implementation of valuation-driven tax strategies. Larry has helped numerous companies professionalize their businesses in preparation for a potential liquidity event, including financial process improvements, vendor upgrades and risk mitigation.

Larry holds an MBA with a concentration in Finance from the University of Chicago and a B.A. in Liberal Arts and Sciences cum laude from the University of Illinois at Urbana-Champaign. He is a Chartered Financial Analyst (CFA), and a past President of the Washington State Chapter of Financial Executives International. Larry has extensive experience teaching financial topics to a variety of audiences.