Skip to main content
search

As rapid, more extreme pivots become the norm, finance and accounting leaders face an expanding portfolio of complex and demanding responsibilities, from value creation and corporate strategy to risk and compliance to talent. To succeed in this dynamic environment, the office of the CFO must take advantage of all available technology and data resources, including artificial intelligence (AI) and machine learning (ML).

Automation, and the underlying AI/ML capabilities that drive it, has become a cornerstone of modern F&A, as future-focused CFOs embrace finance transformation to streamline routine tasks, perform more value-added work, and allow non-technical end users to be more self-sufficient with data and reporting. But the latest advances in AI/ML are quickly enabling new solutions that, until recently, may have seemed like science fiction. Harnessing this power for finance and accounting provides compelling possibilities for solving the age-old CFO challenge of doing more with less and doing it faster.

In this first of a multi-part series, we’ll look at the key components of a successful AI/ML strategy, potential use cases, and what it takes to get started on the path to an AI-driven finance and accounting function.

Key Concepts: Different Branches in the AI Family Tree

Artificial intelligence (AI): This concept describes a computer’s ability to apply typical human cognitive functions such as problem-solving, learning and advanced logic.

Machine learning (ML): A subset of AI that uses mathematical models to analyze data unsupervised and get better at it as time goes on. A common example is pattern recognition for fraud detection, image recognition and self-driving cars.

Natural language processing (NLP): If you’ve ever interacted with Alexa or Siri, you already have an idea of what NLP is. It’s the ability of a computer to process data intelligently using human-based language — both spoken and text-based — as well as unstructured data in the form of emails, contracts, websites and more.

Generative AI: Generative AI is a broad category of artificial intelligence focused on creating or generating new content, such as images, text, audio, and more.

Large language model (LLM): LLMs are a subset of generative AI trained on massive amounts of text data to understand and generate human-like language. ChatGPT is the best-known example — and it wrote this definition!

Embracing AI/ML for Finance

AI/ML is no longer pie-in-the-sky futuristic slideware — it’s already here. So, there’s almost no choice but to embrace AI/ML and accept it as part of modern finance operations. Chances are the typical F&A employee is already surrounded by AI/ML but not fully leveraging it — or may not even know it’s there.

The good news? It’s not as difficult to get started as it used to be.

AI/ML task automation and fraud detection are two of the key use cases for the office of the CFO (IDC, 2023). Repetitive processes, forecasting and reporting activities, fraud and risk tasks, and research-related needs are a great place to start. These areas are typically target-rich environments for AI/ML applications. Also, consider the office of the CFO and your organization’s overall strategic goals and assess how AI/ML can help achieve those goals.

Thanks in part to significant advances in cost-effective computing power and storage capacity, AI-powered learning models can consume vast amounts of data and form a human-like understanding of it. For example, the uber-popular ChatGPT is based on LLM and generative AI models that do precisely that with publicly available internet data to provide human-like responses and output such as text, program code or images based on simple natural language commands – known as prompts.

Building Blocks for AI/ML Success

As with other transformative, disruptive technologies, AI/ML is a journey, not a destination. Risk-averse enterprises often wait to invest in newer technologies until a certain level of maturity is reached. But, based on the speed of innovation and new AI/ML use cases identified daily, the typical approach of waiting for stability will not work. Organizations must engage immediately, understanding that more changes will inevitably come.

Starting the journey does not have to be complex, expensive or overly disruptive. Many AI/ML journeys start with grass-roots departmental efforts on small-scale experiments and grow from there. Here are a few critical guidelines to get started:

01.

Start With the Business

When new technologies are trending, business leaders frequently feel pressure to find problems to solve with the new solutions. While this is a good thing, it’s important to consider how the solution aligns with your overall business strategy and how the outcomes can be tied together. Defining clear use cases with measurable outcomes is a must before embarking on the AI/ML journey.

Many pilot projects using promising technologies fail due to either unclear value propositions or the outcomes cannot be measured effectively. Consequently, the technology is declared a failure and is abandoned. AI/ML is not something that should be abandoned after small failures.

02.

Take Inventory of Tools and Talent

Disruption doesn’t typically come from IT. It comes from the business with IT support. The reality is that many IT departments are struggling to keep up with the AI/ML trends, so they must be creative and develop their capabilities. Fortunately, most of the newer enabling technologies are becoming more user-friendly for those without technical backgrounds. Fintech and ERP systems — even spreadsheets — now have highly advanced data science capabilities that do not require a data scientist to operate.

Look to your power users and other team members who have an affinity for working with the technical side of things. Also, find out how other departments are using AI/ML to get a more holistic view of the capabilities within your enterprise.

03.

It's STILL About the Data

AI/ML only magnifies a problem that has always plagued finance and accounting — bad data produces bad results. In fact, many consider data quality to be the primary obstacle for AI/ML success (iMerit, 2023).

Many organizations still wrestle with poor quality, incomplete, or unavailable data, even if some level of data governance exists. Enterprise data warehouses or data lakes provide data that feeds AI/ML. While this can work for transaction-related data such as accounts payable, accounts receivable, general ledger, etc., it may come up short when overlaid with complex calculations and reporting used in financial applications.

It is also critical to not overlook data security and risks. As tools such as ChatGPT become more powerful, their usage in cyberattacks and data breaches will increase exponentially. Many consider AI/ML as the single biggest emerging cybersecurity threat for the immediate future (Forrester, 2023).

Consider true data needs and shortcomings before you start your AI/ML journey and be prepared to either increase data quality or understand and acknowledge potential traps upfront to manage expectations.

04.

Start Small, Fail Fast, and Build on Success

Don’t wait for the “killer app.” Start now and start small. AI/ML is still very much in its infancy as far as the business world is concerned. It’s changing almost daily and is nowhere near perfect. The possibility that you fail to achieve the expected outcome on some of the earlier projects is very high. It will happen, but you must stay vigilant and not give up. Utilize Agile principles to identify potential failures and course-correct early and often. Once you’ve figured out what works and doesn’t, move on to a bigger project and build on the success.

Tools to Get You Started

In the meantime, identify and quantify a few realistic use cases. Start small and start with what you have — waiting on the sidelines is not an option with AI/ML. Those who do will find themselves drastically behind the curve, playing catch-up in the blink of an eye.

These exercises will be disruptive. And that’s the point. As finance transformations occur and new processes and technologies are introduced, they disrupt the old way of doing business and become more automated and data-driven. Adopting AL/ML is at the heart of this disruption and should be considered for every process today and tomorrow. The future is here.