We’ve come a long way in finance and accounting. From purely manual bookkeeping carried out in large, dusty, paper ledgers, through to Excel-based solutions and advanced accounting systems, we’ve been plunged into the exciting world of intelligent automation – powered by the “golden triangle” of robotic process automation (RPA), artificial intelligence (AI), and smart analytics.
Representing a complete game-changer in improving finance processes, particularly promising under the umbrella of AI, is the concept of machine learning (ML). ML algorithms enable a specific task to be performed without it being explicitly programmed but rather through learning by example. The solution is based on statistical models created from sample data provided to the algorithm. The machine learns from past experience and makes decisions through analyses and predictions while gaining new knowledge from new events.
But how can ML be applied to finance processes, and what is the perfect recipe for a best-in-class ML solution? Here are three basic questions you can consider when developing an ML solution:
What can I improve using ML in finance?
The first step is to select a process that is suitable for enhancement by ML – for example, a process that contains a “thinking” factor such as analysis. Processes that involve only simple data collection, formatting, or calculating a value based on a given, fixed rule are more suitable for RPA. Other than that, the sky is the limit.
The process should also have been repeated enough to have the right amount of existing data for the machine to learn. In theory, previous process performance can also be simulated using data from a company’s accounting system. A learning machine uses this “synthetic” data to learn how to execute the process in the future.
Why should I implement an ML solution?
Second, you need to create a valid and strong business case for building an ML solution. You need to calculate and demonstrate the tangible, fact-based estimated benefits that can be delivered using ML, as well as the amount of human labor you can save. Other benefits of the proposed ML solution might include faster decision-making through carrying out faster analysis, increased accuracy of the output by eliminating human errors, and consistent presentation of your analysis.
The costs of building and deploying an ML solution also need to be considered. These typically include the cost of solution-building, continuous learning from new cases, maintaining a deployed machine, and correcting any errors the machine might make. Understanding the benefits and costs enables you to then create a proof of value for your ML solution.
How do I build an ML solution?
The third step is confirming that your ML solution is technically feasible by ensuring that you have enough data samples and that the data quality is good enough for the machine to learn.
Other components required for an ML solution also need to be defined. For example, does it need a robot to gather and format data for your solution to analyze? And since ML speaks in the language of digits, you will need a “translator” tool to present the results of ML analysis in a format and language that can be understood by human recipients.
Once you understand all the solution details, you can work on defining the exact budget and roles required for the deployment of your ML project. This includes having qualified people in place to check the performance of your solution and help solve any “new cases” that may occur in the future.
The intelligent machine of the future
In a perfect world, our intelligent machine may well be able to learn the rules, standards, and exceptions of accounting from a “virtual university,” and then use this knowledge to support bookkeeping decisions. The majority of the finance department would then be left to work on building and maintaining these intelligent machines, transforming the traditional accountant-based finance function into a team of knowledge-based AI workers with a completely different skill set.
Seems like a distant future? It might be not as distant as you think.
Look ahead to “Future-Forward Finance: Shifts And Evolutions.”
Capgemini is an SAP global partner.