CFOs Guide to AI and Machine Learning by Oracle
Executive Summary
As artificial intelligence (AI) and machine learning (ML) continue to dominate business discourse, CFOs face increasing pressure to adopt these technologies. However, Oracle’s CFO's Guide to AI and Machine Learning emphasizes a measured approach to implementation—one that prioritizes business value over hype. This article summarizes Oracle's key insights while providing additional context on how executives should think about AI in relation to cost, return on investment (ROI), and operational readiness.
AI is powerful but resource-intensive, requiring significant investment in computing infrastructure, data quality, and workforce training. It should only be used in high-value scenarios—such as advanced analytics, dynamic pricing, and pharmaceutical research—where ROI justifies the complexity. For most companies, foundational technologies like Robotic Process Automation (RPA) and machine learning offer more immediate value. These tools automate repetitive tasks and generate usable business data, building the groundwork for future AI initiatives.
Understanding the distinctions between RPA, ML, and AI is critical. RPA is process-driven; ML is data-driven; and AI, especially deep learning, combines both to uncover insights and automate complex decisions. CFOs are advised to optimize automation, establish robust data practices, and scale gradually into AI solutions. Companies that follow this phased approach will be better positioned to leverage AI strategically, rather than reactively.
An intro to AI
Because implementing AI is costly, complex, and often misunderstood, CFOs should prioritize automation and process optimization before adopting AI technologies in their businesses.
Oracle recently published a business guide, “The CFOs Guide to AI and Machine Learning”. Here's a breakdown of the key points and takeaways from Oracle’s guide—and my thoughts on how CFOs can think about AI and machine learning.
The term AI is everywhere right now. It’s being used as a blanket term to describe things that aren’t even AI like simple automation tasks. Many companies are eager to dive headfirst into AI, hoping it will solve pressing business challenges.While I understand this and am excited to learn more about AI myself, we need to know the right use cases for it.
“Microsoft reportedly spent hundreds of millions just to create the [ChatGPT] computing infrastructure where the system was trained and now lives” (Oracle NetSuite, 2024).
This is mainly due to the computing and energy costs. AI needs to be fast to be useful. This demands cutting-edge technology like Cloud-infrastructure and Fiber Optics. Due to recent technological advancements, AI is starting to become a viable business tool. However, while AI has been touted as expensive; DeepSeek changed that perspective and scared AI markets earlier this year, pictured below. DeepSeek is Chinese startup that develops Large Language Models (LLM) like competitors, ChatGPT or Claude. To learn more, read this article on the DeepSeek market scare by Reuters.
NVIDIA (NVDA) Stock Chart
Microsoft (MSFT) Stock Chart
However, it is still yet to be seen if DeepSeek is as cost adverse as discussed or if other factors are going into the pricing structure; possibly, operating at a loss to capture a growing market or using data as value from millions of people engaging on a regular basis and providing lots of personal data. As AI evolves, we will see how different competitors take up space in the market.
Companies like Google and Amazon gain true value from millions of people engaging on a regular basis and providing lots of personal data. It isn’t surprising that a technology replacing Google is possibly worth billions of dollars.
“Alphabet is the world's third-largest technology company by revenue, after Amazon and Apple, the largest technology company by profit, and one of the world's most valuable companies” (Wiki., 2025).
The thought of an LLM like Claude, ChatGPT, or DeepSeek capturing a portion of Google’s market isn’t far from reality.
“Google Search is responsible for 57% of Google’s earning… the 2023 Total Google Search & Other Revenue was $175.04 billion… In 2024, Google is worth approximately $2.04 trillion.” (Bosze, 2024)
Google Search and advertising is a huge portion of revenue for Google. How will ad sales be affected when Google’s usage rate begins to drop? How will LLMs capitalize on this market takeover?
So, when should we use AI?
For the tasks that can’t be solved without AI and for companies that have tons of data on customer behavior and other key metrics; AI can be used to uncover unique insights and correlations not easily found by humans.
Companies aren’t just using AI for the sake of it. Most businesses don’t have the resources or expertise to implement it in the first place. AI is cost and labor intensive, so AI needs to have huge ROI.
AI can be used for life altering tasks especially in healthcare. Pharma researchers use AI to simulate millions of chemical compound interactions to develop new drug therapies. The benefits are there due to their ability to model countless drug interactions in minutes. This is the huge ROI we are talking about. This simply would be unattainable without AI.
There are amazing use cases for AI and the hype is well deserved. Another use case are intelligent prediction systems that analyze the effects of price changes including customer sentiment and sales.
The internet is a game changer for most businesses. Big e-commerce brands have gotten rich off the internet like Amazon but there are also many small businesses that are able to survive and thrive due to the internet. The consumer also wins with ease of access to products. The issue is customers are much more willing to price compare on different sites because of how quick and easy it is.
“Machine learning systems can gather competitive intelligence by scouring the internet and use it, along with your own data on buyer behavior, to determine price elasticity and predict customer trends, sometimes at a personal level” (Oracle NetSuite, 2024).
If you think about 30 years ago, people drove to every store to see the price. Now, you can do this in minutes by visiting a few webpages. This newfound power to the consumer makes it more difficult to determine what consumers will do in the face of price changes.
“Some machine learning systems even help retailers set dynamic prices, maximizing the revenue from goods or services with finite supply, like concert tickets or limited-edition items. What will the market bear for a popup performance in an intimate venue, a hand-painted bathtub, or a 50-year-old bottle of Scotch?” (Oracle NetSuite, 2024).
These examples show some benefits to technology and the new possibilities due to computers, mathematicians, engineers, and developers. We may not be there yet in AI, but people have truly invented life-altering technology. The fact is humans can’t compete with computers in specific tasks. The sooner businesses truly adopt technology and use it to its full potential; the sooner businesses will unlock their full potential.
The information through data analysis and the tools you can use to ease business processes allow for companies to function at a rate like never before. Learning about your markets, products, and customers in a way you had no idea of until you aggregate business data and tell a story. Identifying a business issue, asking why and how to solve it. However, this doesn’t mean dive headfirst into AI without exploring other options. Keep it simple! If a problem can be solved without AI or ML, it should be. Again, AI can be complex so use it sparingly where there are huge ROI opportunities.
Alternatives to AI — Automation and Machine Learning
Before diving into AI, companies should invest in automation and machine learning (ML), both of which offer meaningful business impact with lower complexity. Automation is ideal for repetitive, time-consuming tasks. It improves accuracy, saves time, and frees up resources—making it a natural first step. It follows strict rules and workflow logic, such as sending order confirmation emails or processing payroll. In short, automation is process-driven, whereas AI is data-driven and requires both structured data and workflows to function effectively.
Some common automation use cases include:
Digitizing & classifying invoices
3-way expense matching
Running payroll
Closing books
Once automation is in place, companies—especially large enterprises or those in data-rich environments—can begin exploring machine learning. ML is algorithm-based: we input large volumes of labeled data into a model to produce predictive or classification outputs. It doesn’t adjust itself unless explicitly reprogrammed.
ML is especially effective for tasks with clear outcomes, such as spam detection, where patterns like suspicious links or poor grammar help classify emails as spam or not. ML also excels in reading and classifying documents, digitizing transactions, and unlocking valuable business insights.
Together, automation and ML provide a strong foundation. They allow businesses to improve operations while preparing for more advanced AI implementations in the future.
Enter Deep Learning.
With Deep Learning, algorithms update via training. This AI can be compared to a human brain. The more info we give it, the more it learns about a certain task; the better it will perform.
This type of AI learning is done using artificial neural networks. Just like your brains uses millions of neurons to process information. Artificial Neural Networks use a simultaneous layering approach like how your brain processes the world around you.
A typical computer has 4 to 8 processing cores, while deep learning systems use thousands of GPUs to process massive datasets in parallel. Compare this to a brain that has millions of these “processing cores”.
You can think of cores as thought processes that happen simultaneously. This shows just how advanced our brains are when it comes to processing versus a computer. However, deep learning may quickly become better than a human brain which has the potential for amazing benefits but also serious ethical concerns like unemployment and privacy.
ChatGPT is one example of Deep learning, and companies have spent millions on the architectural capabilities required to support these AI systems. ChatGPT is great for brainstorming and process planning. However, when it comes to taking action that is not what ChatGPT is for.
ChatGPT is the brain, but it needs tools to be able to interact with the outside environment and cause change in the real world. This is where tools like AI agents come into play. We can leverage the deep learning capabilities of ChatGPT coupled with the real-world automation capabilities. This technology is still developing but may lead to the classical AI popularized in movies. For more info on AI Agents, you can read my article titled “AI Agents are the Future”.
Conclusion
In conclusion, with the immense innovation in computing capability, we are finally catching up to the AI theory that has been worked on since 1956. AI and associated technologies are not cheap but starting to become a more viable option for businesses. Even though it isn’t a fix all and companies should start by considering other options before taking on the heavy task of AI.
Nevertheless, AI is an amazing and rapidly developing market. It can benefit humans in many ways including healthcare and business. CFOs and other executives will continue to implement more practical solutions in their businesses and AI is going to be a part of that technology stack. To do this, companies will need good data, automated processes, and knowledgeable people to plan and implement these solutions.
To download and read the full article from Oracle, you can visit this link on Oracle’s website. If want to chat more on this topic, I encourage you to send me a message.
What CFOs Should Do Now
Oracle’s CFO’s Guide to AI and Machine Learning advises finance leaders to take a phased, strategic approach to AI adoption. As the guide explains, “AI should be applied to high-value use cases where ROI justifies the investment,” such as dynamic pricing models (Oracle NetSuite, 2024). This caution stems from the fact that AI requires substantial infrastructure, talent, and clean, organized data to succeed.
From my perspective, CFOs should first focus on automation and machine learning before fully diving into AI. Technologies like robotic process automation (RPA) and ML can deliver meaningful results - such as automating invoicing or detecting expense anomalies - without the heavy lift of full-scale AI. These tools not only streamline operations but also help companies build the data maturity required for future AI initiatives. By strengthening their automation and data foundations now, CFOs will be better equipped to evaluate and deploy AI where it matters most.
Work Cited
Bosze, A. (2024, October 15). Google revenue breakdown (2024). Doofinder. https://www.doofinder.com/en/statistics/google-revenue-breakdown
Carew, Cooper, & Banerjee. (n.d.). Deepseek Sparks AI stock selloff; Nvidia posts record market-cap loss | Reuters. DeepSeek sparks AI stock selloff; Nvidia posts record market-cap loss. https://www.reuters.com/technology/chinas-deepseek-sets-off-ai-market-rout-2025-01-27/
Google. (n.d.-a). Microsoft Corp (MSFT) stock price & news. Google Finance. https://www.google.com/finance/quote/MSFT:NASDAQ?sa=X&ved=2ahUKEwjTs-fA2dqNAxUyFlkFHe2XJgoQ3ecFegQIPhAb
Google. (n.d.). Nvidia Corp (NVDA) stock price & news. Google Finance. https://www.google.com/finance/quote/NVDA:NASDAQ?sa=X&ved=2ahUKEwjvjs311tqNAxXtEVkFHVVGAfQQ3ecFegQIPhAb
Oracle NetSuite. (2024). The CFO’s Guide to AI and Machine Learning.
Wikimedia Foundation. (2025, June 2). Alphabet Inc.. Wikipedia. https://en.wikipedia.org/wiki/Alphabet_Inc.
