Top Use Cases Driving Adoption of AI/ML in Finance & Internal Audit

Regardless of the hype and the “fear of missing out,” today’s finance leaders are under pressure to achieve better results faster and with limited resources. AI/ML offers a variety of ways to streamline and innovate finance and accounting as well as internal audit, as we outlined in the first article of our multi-part series. Here, we explore some of the top use cases and the underlying AI/ML features that enable each of them.

Key AI/ML Capabilities and Their Financial Use Cases

Breaking down modern technologies into key capabilities can form a basis for developing use cases for your own journey and success stories.

Prediction and forecasting

Financial leaders are expected to predict the future in many cases. For a long time, we’ve prepared budgets and estimates based on historical data using simple methods and tools such as spreadsheets. Today, advanced AI-enabled forecasting models coupled with significantly more input variables, data, and computing power have enabled scenario planning that goes far beyond simple historical data.

Developing a revenue projection based on historical sales data was the common approach forever, but by including external data such as weather, economic forecasts, pending regulations, and other non-traditional data sources it’s possible to develop a much more reliable and comprehensive future picture. Doing this without the help of AI/ML and advanced cloud computing would be prohibitively time-consuming or even impossible. Popular use cases include:

• Demand and revenue forecasting
• Financial scenario planning
• More accurate pictures of future-state asset valuations

Unstructured data analysis

Structured transactional data drives the day-to-day business operations of most companies. While that same data forms the basis for many finance-related calculations, there is another set of critical data that resides in contract documents, document images, and linked spreadsheets as well as other data that does not typically come from financial ERP systems.

Natural language processing (NLP) and large language models (LLM) are two AI/ML technologies that “learn” from large amounts of unstructured documents and create a human-like interaction capability where users can get answers by asking simple questions.

For example, NLP and LLM can precisely summarize large, complex accounting regulations in simple terms — almost instantly. This can also be useful for summarizing large contracts into simple, salient terms that are easy to understand. Popular use cases include:

• Compliance and regulation interpretation and summaries
• Streamlining management and usage of complex spreadsheets and other datasets
• User-friendly chatbots to query internal financial data
• Contract management and usage

Pattern recognition

Identifying and preventing fraud is important to every financial professional, but this is easier said than done. Historically, to capture fraud you had to specify what to look for along with various permutations of the fraudulent activity. Bad actors who commit fraud are continuously refining their methods, which makes it even more difficult to catch.

Pattern recognition helps to level the playing field by continuously monitoring transactions and looking for suspicious patterns that may not be visible by traditional human analysis.

Consider vendor payment fraud. Typically, you might monitor data fields such as addresses, amounts, and vendor names to look for known fraudulent patterns. AI/ML can learn from these known patterns and adapt to similar patterns and detect anomalies that a human might never have detected, such as payment timing, goods or services being procured, or suspicious buyer behavior.

This continuous monitoring and learning can significantly increase the chances of catching and preventing fraud. Common use cases include:

• Fraud detection and prevention
• Customer experience and acquisition
• Process optimization

Task automation

Financial professionals must work efficiently and effectively to complete the seemingly never-ending plate full of projects they have before them. Transforming your finance function to focus on value-added, knowledge-based tasks is a far better time investment than completing repetitive manual tasks such as entering data or creating reports.

Robotic process automation (RPA) has been around for a while and has become popular for repetitive tasks that do not require complex logic or functional knowledge. Similarly, purpose-built platforms such as BlackLine can automate a lot of transactional work, liberating your workforce from mundane but necessary tasks.

AI/ML is making these tools smarter, so instead of simply plugging data into screens, they can now make financial-based logic decisions to increase their accuracy and productivity. New generative AI tools such as Microsoft Copilot for Office365 automate the creation of complex spreadsheets and graphs simply by describing what the user wants to see. It’s no longer necessary to spend time learning the formulas and syntax — just tell Copilot what you want to see, and it does the work in a matter of seconds. Use cases include:

• Intelligent data entry
• Document approval and workflow
• Financial statement preparation

Planning for AI/ML Opportunities and Implications

These are just a few examples based on the underlying capabilities available in today’s AI/ML tools. Although the possibilities are limitless, it’s important to keep in mind that — like everything technology-related — there is no silver bullet. Every solution comes with certain drawbacks and requires a combination of factors to succeed.

As you set out on your own journey, some of the factors you need to think about include your corporate culture, enterprise risk profile, and the quality of your data. Carefully consider how your organization will control the growth and proliferation of AI/ML solutions across the firm.

Tools are getting more powerful and easier for non-technical users to create solutions with. Without adequate controls, the situation could quickly get out of hand — to a point where instituting any control may be difficult.

The same goes for data. AI/ML solutions will only be as good as the underlying data, which we all know is an age-old struggle. AI/ML will only amplify this problem as more solutions are deployed and get more attention from leadership. You only get one chance to make a first impression and the hype around AI/ML is deafening as the critics take notice. One failure out of the gate could poison what otherwise could be a game-changer for your firm.

Upcoming articles in our series will dive a bit deeper into these factors and share ideas on how best to prepare for a more successful AI/ML journey in finance. Subscribe to our Now of Work newsletter to stay up to date.