By Kenton Chow
Artificial intelligence (AI) references are everywhere, and AI is being discussed by everyone from Elon Musk to Henry Kissinger. AI combines vast sets of data and computer science to efficiently solve problems and utilizes machine learning to leverage pattern recognition based on those datasets. AI is the new technology buzzword, but it is no longer just conceptual; it’s now in widespread use. In fact, AI has been emerging for many years and now represents the fastest technology adoption ever and is now mainstream from Microsoft Azure GPT-4/ChatGPT to, Google Bard to Open AI.
Have you ever searched a topic on-line using the Google search engine and were presented several different options to view? Of course you have. Search algorithms are one of the earliest examples of machine learning based AI. There are countless other AI use case examples and AI experts, researchers, businesses, and governments are all engaged in active discussions about how AI will impact every aspect of our lives.
This article will focus on the challenges that AI represents within the Finance function and best practices for CFOs looking to incorporate AI into their organizations.
Benefits of Adopting Artificial Intelligence in Finance
AI represents an enormous opportunity for CFOs to add unprecedented efficiency and effectiveness to the operation of their organizations, which can significantly reduce cost and improve profitability.
One of the more straightforward opportunities associated with AI is to utilize AI tools to enhance process efficiency related to repetitive tasks, such as data entry and customer queries. For example, Delta Airlines uses AI-based chat tools to answer customer questions that formerly were answered by a customer service agent over the phone. By integrating AI into back-office operations using tools like ChatGPT, organizations can streamline processes, increase efficiency, and improve customer satisfaction while (importantly) also lowering costs. And as AI creates opportunities to improve performance by increasing human productivity, Finance can increase its strategic focus on the business.
Dynamic Pricing and Customer Insights
An area where AI-based tools may become even more interesting is by providing financial analysis and insights. How much impact to revenue growth could be obtained if CFOs had optimal predictive surge pricing based on dynamic demand characteristics? The event ticketing and the airline industries have already adopted AI tools to dramatically improve revenue optimization by enabling dynamic pricing based on demand (e.g., for Taylor Swift concert ticket prices or airfares over the July 4th weekend). Predictive pricing to optimize revenue is nothing new, but AI tools that are based on machine learning make pricing far more efficient than it has been in the past. For example, e-commerce retailers are excellent in Knowing Your Customer (KYC). They run promotions and discounts to a very targeted set of consumers by knowing their customer’s specific preferences. AI-based predictive data analytics can then provide insights into customer buying behaviors based on various price points to optimize what product or service is offered at what price point and at what time of the day. These powerful customer insights typically result in much more optimized revenue and profits.
Forecast visibility is incredibly valuable to a CFO. If data is highly predictable, that data-driven insight is indispensable, but this becomes a big challenge if the data is unreliable and inconsistent, resulting in unpredictable outcomes. We are now seeing Wall Street analysts use AI tools to build financial models and projections more efficiently for public companies. In the past, staff analysts would build financial models from scratch to assist in providing some “visibility” into future projections of company performance. What AI-based tools do is to help provide more predictable, higher probability outcomes, and hence a more accurate set of projections and greater forecast visibility. At some point AI-based tools may start to develop financial analyst reports from Wall Street with far less human input.
Artificial intelligence tools are also being utilized by companies to improve their operational planning. One of the key responsibilities of a CFO is the allocation of resources to drive the most financially efficient results and profitability. Having tools that can predict where the most efficient use of capital is, helps to optimize return on investment when deploying scarce capital resources. These insights lead to better operational decision making, such as how much headcount to hire and how much inventory to purchase.
Risk Mitigation: Fraud and Embezzlement
The last benefit associated with AI I would like to address is with respect to fraud and embezzlement detection. This is a huge opportunity for financial institutions and retailers, among others. Financial transaction fraud is so prevalent that it has become an epidemic. AI enhances fraud prevention systems by improving the accuracy of predictions identifying which types of transactions are fraudulent. By holistically analyzing historical data from past fraudulent transactions, supervised machine learning models can be trained to identify fraud in future transactions.
In cases of new fraudulent behavior, unsupervised models, which do not require historically labeled data, AI tools will also flag abnormal transactions that show indications of potential fraud. These flagged transactions can be sent on to manual reviewers for further investigation. Newly discovered fraudulent transactions can then be fed back into AI machine learning models to retrain these on an ongoing basis, allowing them to continuously adapt to the latest patterns and behaviors of fraud.
Overall, this means financial institutions will reduce the number of fraudulent transactions that go undetected, as well as avoid disruptions in the customer buying experience by preventing genuine transactions from being flagged incorrectly, which is worth billions in cost savings and customer goodwill.
Risks of Adopting AI
Although the benefits of adopting AI are considerable, there are risks and downsides as well. Currently, Al has several limitations for its practical use in Finance. This is because algorithms, such as ChatGPT, can only answer questions that they have been trained to answer. Without the detailed data of a CFO’s specific organization, AI typically can’t provide accurate feedback at the company level. Moreover, ChatGPT will try to provide an answer to a query even if the AI model doesn’t have the data and therefore ability to answer it correctly.
Job Shifts and/or Job Losses
Loss of jobs are a significant concern for many businesses, and AI has the potential to create both job loss and job shifts. Most people employed in Finance will tell you there is no shortage of work to do in that function. In many cases, Finance is just barely able to keep pace with the bare minimum of running the business and meeting regulatory requirements.
AI may ultimately cause many jobs to shift in their responsibilities. Some roles will be lost, and some new jobs will be created. AI won’t replace people per se, but it will change people’s roles within a company anywhere that AI tools can offer a more productive way to optimize operations. So, rather than seeing automation in the Finance function reducing the number of staff, for example, CFOs should look at it as a way to increase the scope of what the function can achieve with the same headcount.
AI and Crime
Another risk of AI is that it becomes an effective tool for criminal and illegal use. For instance, data security is a topic on the minds of many company executives, federal program administrators and national security experts. Security breaches and hacks of customer data, personal information and even defense systems are big concerns. Data security has long been a significant area of risk, but since AI tools are based on large amounts of datasets and databases, access to data is also a prime target for cyber criminals and hackers. The use of AI-based tools to illegally identify vulnerabilities in data networks, such as Amazon Web Services, Google Cloud or Microsoft Azure, could be catastrophic to the global economy. Billions of dollars could be made by cyber criminals through ransomware around the selling of intellectual property and/or personal data. For example, the Colonial Pipeline data network was breached in 2021 and was faced with ransomware, causing oil supply industry disruption and requiring the payment of a hefty ransom.
Lack of Industry Standards, Regulation and Protections
Both our corporate and governmental leaders have a responsibility to establish checks and balances for the use of AI. Without human ethics and judgement, AI researchers worry that AI-based machine learning could blindly execute actions based on the training the machines have accumulated. If the US national defense systems were breached, for example, AI models could cause international conflicts or even war without human judgement involved. This is where policy, regulation and protections must be established in the appropriate and safe us of AI technologies.
Built in Bias in Data Sets
The last AI risk to highlight is that machine learning based tools are inherently biased and prejudiced by definition because AI is only as effective as the data the machines are being trained upon. If datasets are biased, then AI results will likely reflect the same bias as the historical dataset. One example that comes to mind are local, state and national voting records. Biased voter datasets can significantly impact the information disbursed to targeted voter groups with both intended and unintended consequences for government. Another example is the pricing of consumer loans based on political affiliation race, gender, and region. There has long been historical pricing prejudice in consumer loans, and AI could institutionalize that bias because of reliance on these historical data sets, perpetuating the same historical outcomes.
It’s clear that AI tools are evolving and being adopted at a rapid pace but are associated with the above risks and benefits.
So, what is a CFO to do? Here are some AI best practices for CFOs to consider.
Evolving AI Best Practices for CFOs
Practical AI Use Assessment in Finance
One key best practice is for CFOs to select a very narrowly defined task for any AI application, after which humans need to examine the application’s outcomes and make changes to the processes and machines involved, as errors and exceptions will arise.
AI has some limitations for practical use in Finance. CFOs must prioritize prerequisites for future use of ChatGPT and similar technologies within Finance, including identifying appropriate potential use cases and discussing how AI tools and other large machine learning based models could fit within Finance’s technology strategy, as well as how employees view this technology.
For example, using ChapGPT to help a Finance leader generate software code or SQL statements is a great use; however, asking it questions about how to interpret a specific company’s audited financial statements won’t work . Eventually, we can expect to have secure ChatGPT-like services that are customized to specific organizations, but this will take time. For now, leaders should focus on transitioning a company’s culture and skills towards using advanced technologies such as AI and make sure that their teams have the right capabilities as technologies advance.
Establish a Corporate AI Policy
CFOs can take the lead in establishing an AI policy for their company by providing guidance for the organization about what is and is not an appropriate use of AI tools within the enterprise. For example, Samsung has banned certain AI tools after discovering that some of the software source code had been shared in the public domain through an AI tool. CFOs should use their unique risk-focused perspective to push for, and help, shape a policy at the organizational level that mitigates the risks posed by unrestricted use of AI tools, including output quality risks, data security risks, and regulatory risks, as well as their downstream implications.
Conduct a Compliance and Confidentiality Review
CFOs can also spearhead a compliance and confidentiality review evaluation across the company what is confidential and what isn’t. This would include the process for determining what is confidential and what can be shared and not shared.
Provide Guidance for AI Model Use Across the Enterprise
Finance leaders can help provide structures for governance of AI models to help ensure end-to-end AI governance, from the definition of a company’s AI strategy, through training, testing, deploying, and monitoring of AI across departments and the data used to train models. My advice here is to stay flexible because regulations and legal guidance are always lagging. For example, AI tools like ChatGPT are being adopted faster than most corporate governance rules can be established. With a lack of external guidance in these situations, company leaders will be on their own to make their own judgement calls to determine when AI boosts efficiency and when it doesn’t.
Artificial intelligence tools represent a tremendous opportunity, but CFOs must really think through the benefits versus the downside risks when integrating AI tools into their organizations. Proactive AI assessment by CFOs can help integrate AI more successfully within the enterprise and focus these technologies to specific, appropriate tasks while recognizing that humans must then look at the outcomes and make changes as errors and exceptions arise. Over time AI will change nearly every aspect of our lives, so we all need to be prepared to embrace AI as it continues to rapidly evolve.