Understanding Machine Learning For The Digital Finance Function

Machine learning fits into the ‘Intelligent automation’ technology pillar as part of the Digital Finance Function Model (DFFM), in which it is paired with Robotic Process Automation  (RPA) to digitalise processes. According to McKinsey, it could automate 64% of data collection and 69% of data processing roles.

What is machine learning?

Machine learning is a form of artificial intelligence. Essentially, it’s an algorithm with the ability to draw predictions or conclusions from a historic data set. With the sheer volume of data now available to organisations, machine learning is becoming more and more useful in sorting through that data.

As AIs go, machine learning is usually fairly simple – known as a ‘weak’ or ‘narrow’ artificial intelligence, it can perform tasks within a set of boundaries and rules. For accountants, it can be used to augment core processes to increase efficiency and improve decision making. You could automate repetitive manual tasks and remove ‘swivel-chair processes’.

How you could use machine learning

• Intelligent bookkeeping
• Process suggestions
• Process decisions
• Forecasting and predictions
• Fraud (outlier) detection
• Non-financial (qualitative) predictions
• Natural Language Processing (commentary)

Considerations for machine learning

It its report, Machine Learning: More Science Than Fiction, ACCA’s Narayanan Vaidyanathan outlines five ethical challenges posed by machine learning.

Dealing with bias: There has been much in the news about machine learning programmes that reflect unconscious biases in society – face recognition software that can’t recognise black faces, for example. “This is one of the biggest ethical challenges for ML,” writes Vaidyanathan. “The algorithms, both supervised and unsupervised, may need to be properly interpreted in order to avoid confusing correlation with causation.”

Strategic view of data: It’s no good applying machine learning to data if your data isn’t high quality and well-organised in the first place. “In order to take advantage of data in a sustainable way, an organisation needs a coherent data strategy.”

Assigning accountability: Can you trust your machine learning algorithm to make decisions for the organisation, or should a human be accountable for filtering and checking those outputs?

Looking beyond the hype: With so much buzz around AI, people can develop unrealistic expectations. Technology vendors can also oversell what it can do, so you must be wary when choosing software and planning for transformation.

Acting in the public interest: “Technology can raise universal questions about public good and public value and professional accountants may find themselves being pulled in different directions as a result,” says Vaidyanathan. “Defending the public interest requires an ability to go beyond the basic minimum that is required for legal compliance.”

The skills you need within a machine learning environment

Current machine learning tools, while they can streamline processes and free up a lot of time within a DFF, need human supervision and oversight to really draw the true value from them. Accountants should look to upgrade their tech knowledge and know the platforms they are using inside and out, while also developing core human skills, such as analytical skills, strategic planning, communication, collaboration and relationship building.

Machine learning is part of a suite of technology applications that will form a critical quadrant of your Digital Finance Function.