Finance has been among the earliest and most significant adopters of Information Technology. With the amount of data handled for planning, risk analysis and wealth management, its processing, reporting and analysis have become instinctively data-driven. The sector has adopted innovative technologies to automate, bring transparency, reduce human intervention, eventually eliminate the risk of errors and reduce costs.
Yet, finance teams face multiple roadblocks in providing accurate, timely reports and insights to support business decisions. A big challenge is to bring transparency to the leadership about how each supporting KPI is calculated, and the factors under consideration along with their weightage. Moreover, 80% of financial planning and analysis data pipelines are performed manually with the help of SQL jockeys, using offline databases or multiple spreadsheets.
How can CFOs achieve a technology leap to address these gaps, by not just bringing automation but also bringing intelligence to their function?
Towards more autonomous finance with ML
Machine learning (ML) applications are a natural fit for financial planning and analysis (FP&A) processes. ML is particularly useful given its capability of identifying patterns, trends and providing early predictions of deviations. Given the intricacy of volume, variety and veracity of data that financial planning teams need to analyze, it becomes too challenging a task to be performed by people alone, unassisted by self-learn-technology.
CFOs can use ML in many ways. A practical application of ML is in automating financial data pipelines and data management tasks, like preemptively bringing, caching and processing relevant drill-down data into downstream databases. This can include extracting relevant information from unstructured documents like trade sheets, contracts or even emails. This is remarkably helpful and makes the system fast, audit-friendly and eventually iterates into a reliable autonomous system.
The second important use of ML is in identifying high-risk accounts, predicting devaluation, payment defaults or fraudulent patterns, allowing preventative measures before losses occur. This is true risk management. Thirdly, ML algorithms can be used in wealth management to optimize investment strategies by analyzing market trends and identifying undervalued assets.
ML can also serve as a useful tool for financial modeling and forecasting. When trained on the right data, ML algorithms combine an ultra-wide range of variables-including real-time data-to deliver quicker, more accurate forecasts. This results in better plan monitoring, updating and correcting financial decisions.
Obstructions enroute
While ML has the potential to transform FP&A, some challenges should be considered and addressed to ensure that the process is reliable and ethical. FP&A data can be complex, siloed and multi-directionally structured. The requisite data needs to be sourced from different systems, making it difficult to integrate, co-relate and prepare for training ML algorithms for accurate predictions. Data preparation can be time-consuming and resource-intensive, requiring significant expertise.
ML models can be biased and unfair, leading to discriminatory outcomes. For instance, if the historical data used to train the model contains biases, the model will learn and perpetuate these biases in its predictions. It is crucial to ensure that the training data is diverse, representative and free from any prejudice. Additionally, when the training dataset is small, there is a risk of overfitting, leading to poor results with new, unseen data. These can be addressed by using cross-validation and regularization methods.
ML models can be challenging to interpret, making it difficult for financial professionals to understand how the predictions were arrived at. This can be a significant concern as transparency and explainability are important for trusting the model and incorporating its results into decision-making. Extensive audit-trail logging in the ML process will help track the lineage of data and decisions in a prediction cycle.
While the challenges associated cannot be overlooked, the potential benefits of adoption of ML in FP&A are significant. CFOs are increasingly adopting these technologies to improve their forecasting velocity, risks identification, proactive action and resource efficiency.
The emerging trends
CFOs are prioritizing investment in platforms that support the induction of ML. Replacing legacy systems, which are not driving towards modernization, is the key task of most CFOs.
A technically strong, open and modern data platform that implements desired ML flows shall help CFOs in pushing a drive towards autonomous finance. FP&A teams are also pushing for the use of semantic models to improve data quality and transparency. Data lineage solutions promise to better explain data’s treatment and improve its transparency for decision-makers. Data and ML platforms are building tools and interfaces that target decision-makers as their users, shifting away from an earlier focus on data analysts and techies. The use of a universal semantic layer is emerging as a business-friendly solution for integrating data silos into one go-to source of truth, as a layer between data lakes and visualization tools.
The road ahead
CFOs recognize that adopting data-driven intelligence in FP&A is essential for staying competitive. Technology teams will increasingly play a bigger and futuristic role in identifying the right platform with the right capabilities. Finance leaders must persist in working closely with IT, data analysts and other stakeholders to identify the right technologies, build the necessary capabilities and address any obstacles along the way. Then, let the modernization train roll.
##
ABOUT THE AUTHOR
Rajesh Murthy, COO, Kyvos Insights & Intellicus Technologies
Rajesh Murthy is the Chief Operating Officer at Kyvos Insights, an analytics acceleration platform, and Intellicus Technologies, a leading data analytics & BI platform.
Rajesh is widely recognized as an expert in delivering transformative and scalable data analytics solutions that drive business growth and build a data-driven culture. As a seasoned professional in the domain, he is passionate about empowering companies with the insights they need to make better business decisions.
