Prediction is the killer app for AI in finance.

A CFO’s most important task is to give the CEO and the markets a clear picture of what is likely to happen next quarter, next year, and as many years in the future as possible. AI can help people improve the speed and quality of their predictions.

But there’s no single process that powers prediction; it is the culmination of output from numerous processes used to monitor and manage financial performance. By using AI to improve these processes, predictions will get better, too, according to research from SAP Insights.

However, few processes can be completely automated with AI—people will need to keep an eye on things. When deciding where to apply AI, better results come from identifying the specific tasks within a process that can benefit from it. Tasks with at least one of the following characteristics—and preferably all of them—are the best candidates:

• Density. Poring through complex documents can tax employees’ attention spans. A report from McKinsey & Company relates how a global consumer packaged goods company uses AI to create a draft management discussion and analysis report from its financial data for monthly operational reviews, freeing up finance staff to focus on tasks that require more judgment, such as calculating risk.
• Duration. Some tasks don’t require a lot of mental effort; they just take people forever to finish. Think about how long it takes just to collect financial data for analysis. The European Central Bank uses AI natural language processing models trained with supervisors’ feedback to gather data—news articles, supervisory assessments, and internal documents—about the banks it oversees, and to categorize that data in seconds. Supervisors now spend their time analyzing the information instead of searching for it.
• Defects. The Bank of Canada built a machine learning tool to detect anomalies in regulatory submissions, according to McKinsey. The AI’s daily foraging catches things people wouldn’t. Employees can focus on analysis.

By applying these criteria when considering how to use AI, CFOs can improve the processes that contribute to their group’s predictive powers. And when they make better predictions, they can get better at helping their colleagues across the business to spot new opportunities and avoid risk.

Now that’s a killer app.