Machine Learning: An imperative for Finance upskilling?

I regularly write about how the latest trends and digital disruption are reshaping the Finance functions and businesses of the future. Follow me on LinkedIn or DM to discuss how your work or organization is impacted by these disruptions.

In one of the latest publications in 2023, Gartner has predicted that by 2026, 40% of the roles in the Finance workforce will be reshaped and hence will require an entirely different skill set than what they possess today. They also predict that by 2030, 50% of organizations will be using AI/ML to carry out their bottom-up forecasting and thus resulting in “autonomous operational, demand, and other types of planning” (Source: Gartner Finance Predicts).

One of the roles that will be an integral part of any team, including Finance, would be that of a Data Scientist. One who will be able to help build and deploy Machine Learning models to drive process automation.

But, would that mean the role of Finance professionals, as it stands today will be taken over by Data Scientists and Data Analysts? No one in Finance will have any idea about how these ML models are built and deployed. Does that mean the role of Finance will be restricted to an end user relaying the output they are getting from an ML model or AI engine?

This reshaping of the finance workforce or ways of working, to a more automated, data-driven, and self-serve operating model would require us to flex a different muscle. In short, being a presumed data steward of the organization and a co-pilot to the business strategy, Finance will need to up its game on being Digitally Literate.

In this article, I will explain what it means to be digitally literate in the specific context of AI & Machine learning, and why it is imperative for Finance professionals to stay relevant in the fast-changing world around us.

Let me give you an example.

For an ML algorithm to be deployed effectively, it is critical that it has been trained on a set of training data and then validated through a set of test data.

A training set is a set of data on which the model or algorithm is trained and should be representative data containing enough variables and features. The model is then tested on ‘test data’ to assess its accuracy and performance.

Imagine your company hires a data scientist to create an ML algorithm for detecting anomalous transactions in the General ledger. While Data Scientists will be able to create that code probably within a few hours, the biggest risk it carries is around the ‘training set’ they use, and how it is tested on the ‘test data’.

If that model is trained on a training set that already has a bias towards treating certain transactions as anomalous due to:

• either the training set is too small, or
• does not contain the appropriate variables or features that are representative of the population in the General ledger

you end up having a model which is called an ‘overfitted’ algorithm.

From there, it will never give you an accurate prediction of the results of test data or real data. The results can be more biased towards classifying transactions as ‘anomalous’ rather than ‘normal’. This will not only increase your work but most likely defeats the purpose of deploying an ML model.

That is where the role of a Finance professional comes into play.

Applying the correct domain knowledge on ensuring the training set represents the right subset of the population, will enable the most optimum deployment of ML.

You will work closely with the Data Scientist to ensure the model is neither overfitted nor underfitting. Yes, you don't need to know to code, but just have enough operational knowledge and how different ML algorithm work, to be able to confidently have a discussion on the right goals, objectives, training sets, and outputs of the model.

You should be able to challenge the assumptions and apply professional skepticism to the input and results of the algorithm, rather than blindly accepting what the technical expert tells you. Algorithms have long been getting a bad rep of being ‘black boxes’. However, by applying the right amount of critical thinking, domain knowledge, and digital expertise, you can decipher and possibly use that to the advantage of the organization you are working for.

That is just one example of how Finance domain knowledge, combined with the right amount of digital literacy can exponentially generate huge benefits for the organization.

Opportunities are limitless in this domain and it all comes down to our desire to learn continuously and upskill ourselves to stay relevant. We have a number of free resources available over the internet and all that is needed is a commitment and passion to develop oneself.

Share your thoughts in the comments section below.