DNA of A Digital Finance Function (Part 1)
Generative AI is powerful but often overhyped in finance transformation. This article argues traditional machine learning still drives core value by solving inefficiencies like forecasting and process automation. It emphasises balancing both approaches, using generative AI as an enhancement layer to improve insight, analysis, and decision-making across finance functions.
Originally published in LINKEDIN
Part 1: Avoiding Generative AI Pitfalls
I recently had the opportunity to record a podcast with Glenn Hopper, AI thought leader and host of FP&A Today. Complete podcast of the same name is available on DatarailsYouTube channel. Link in comments.
In this article, which is first of a two part series, I will leverage some key insights that I shared in this podcast along with additional perspective on how to build DNA of a digital finance function.
Role of Traditional vs Generative AI
In one of myprevious LinkedIn articles, I discussed the Generative AI fallacy — a dangerous trajectory where gen AI is shifting our focus away from the benefits and applications we can achieve from traditional machine learning. Before explaining further, let's first understand the difference between the two.
Traditional AI, or machine learning, operates in a more deterministic manner. It learns from historical data to predict specific outcomes, such as identifying fraudulent transactions, segmenting customers based on their buying habits, or forecasting cash flows.
In contrast, generative AI focuses on creating new content, still learning from loads of historical data — but for a different purpose. While traditional AI might predict cash flow for a future period, generative AI can build upon that by offering recommendations based on its vast training data, providing summaries, or conducting variance analyses.
The core of finance transformation, however, still relies on and can unlock massive benefits from processes powered by traditional AI, such as cash flow forecasting, customer segmentation, and process optimization. Generative AI then enhances these foundational elements by enabling more interactive and insightful ways to work with the results. Generative AI is essentially an added layer that enables two-way interaction with outputs of traditional AI.
The reason I advocate traditional AI having more application and value for finance is that, according to some studies, 40-50% of finance teams' time is still spent on manual transaction processing and inefficient processes. Focusing on building machine learning algorithms that address these core challenges is crucial before layering on the complexities of generative AI.
Why the Rush to Generative AI Can Be Detrimental?
Generative AI, with its ability to generate new content, is undeniably captivating. We can now interact with AI in a way that feels more intuitive and engaging. However, this allure can lead to a dangerous pitfall: diverting attention and resources from the foundational work needed to digitally transform finance. While it is a powerful technology, the underlying processes are too complex to let us unlock its full potential.
Think of it this way: If traditional AI is a single-cell amoeba, generative AI is a multicellular organism – still far from being the all-powerful solution we might imagine. Overemphasizing generative AI before addressing core process inefficiencies can lead to increased technical debt and projects stuck in "pilot purgatory."
Solution: The Power of Convergence
While there are some pitfalls, we cannot overlook the massive benefits generative AI can bring to the table. Its sheer ability to generate new content and analyze vast amounts of unstructured data makes it an immensely useful tool for finance applications ranging from financial planning to M&A and ESG. We need to balance the AI tech stack by determining which problems are better suited for traditional AI, like process optimization and forecasting, and how we can converge the generative power of AI on top of these.
For example, in M&A, AI is already streamlining everything from target identification to due diligence and post-merger integration. Using generative AI tools is fast-tracking the contractual and legal analysis, risk identification, and even developing customized integration plans. This frees up finance professionals to focus on strategic decision-making and value creation.
ESG is another area ripe for AI disruption. When converged with techniques like Retrieval-Augmented Generation (RAG), generative AI can analyze complex data sets to help companies measure and report on their ESG performance, identify risks and opportunities, and ensure compliance with evolving regulations. This is crucial for building investor trust and demonstrating a commitment to sustainability.
The key here is to embrace technologies strategically, focusing on solving real business problems as opposed to yielding to the hype. At the end of the day, we should ask the question: is a certain technology making us a better business partner, improving operations, or enhancing employee/customer experience? If not, we can safely look elsewhere.
Stay tuned for Part 2, where we'll dive into the human-centric aspects of digital DNA, the crucial role of continuous learning, and the skills finance professionals need to thrive in this evolving landscape.