Artificial Intelligence – Debunking for Financial Planning
Artificial Intelligence and Machine Learning are phrases that are growing in popularity in the financial world. For many, they are met with an element of trepidation akin to a Charlie Brooker-style dystopian world whereby the robots rule the humans. The reality is that the vast majority of us use elements of Artificial Intelligence (AI) every single day, be it the predictive text on our phones, or suggested items when purchasing online. It needn’t be feared, but its capabilities celebrated.
What Is It & How Does It Work?
Machine learning is a particular use of Artificial Intelligence. For simplicity, we will refer to Artificial Intelligence (AI).
AI is often met with scepticism and general angst, and a degree of mystic. The reality is actually quite straightforward. AI uses a computer’s ability to handle large datasets to identify patterns that are too complex for humans to easily see.
A common misconception is that AI is used to predict the future. In fact, it’s the ability to identify patterns in historic data that is the power of AI. Those patterns are learned by an AI engine and can then be used to sense check or enhance your Forecasting and Budgeting.
Where Does Artificial Intelligence Fit In Financial Planning?
The application of AI in financial planning opens the door for the use of varied, complex, and large data to identify patterns and factors that impact upon a plan. This practically means that you could drive your forecast with the creation of automated complex predictions, which are not possible by traditional modelling approaches.
An overly simplistic example might be predicting ice cream sales. By feeding historic data relating to sales of a range of ice cream product sales, and prevailing weather conditions, AI could identify a precise relationship between the two. It would then be possible to create several forecast scenarios “what if we have a hot/cold or wet/dry summer”. AI could give a very accurate range of demand across products and sensitivity analysis, using historic patterns to guide your forecast.
AI and Financial Planning are becoming more and more acquainted with one another. Indeed, a Gartner survey believed that over half of the finance functions it asked, are currently availing of it.
What Are Some Of The Benefits?
Practical uses of AI in Planning fall into two broad categories – Accuracy and Time-Saving.
With additional data comes additional noise. Bigger files, bigger models, and a greater risk of the FP&A team missing a material inaccuracy. Machine learning allows you to consider factors and accommodate complexity that it simply would not be feasible to model by traditional means. The Ice cream demand example would be one of these.
Where this really starts to get interesting is when a AI algorithm can be deployed and continuously improved in use, in doing so actually “learn” as new data and inputs are translated in the new scenario plans, further advancing the accuracy of the model.
AI allows finance teams to use techniques such as anomaly detection to identify misnomers by calculating an expected value and flagging those that fall outside of a predetermined range. AI could detect the result of an erroneous journal, incorrect invoice value, a user typo while entering their operating plan, amongst many other things.
It could also sense check a sales forecast as being outside expected ranges (up or down!) and alert a user in real-time as they are inputting.
We can make robots that learn from mistakes, yet man still makes the same ones over, and over.
Anthony T. Hincks
At its most rudimental level, AI’s ability to improve accuracy is one of its most prominent and easily accepted benefits for finance teams.
An issue where AI can support FP&A and the wider business on is bias. We are all guilty of finding data to support our assertions but what if we are missing out on a more meaningful and lucrative trend as it hasn’t been considered? Unearthing a greater number of trends whilst removing data prejudice using AI could be the competitive advantage the business needs. We’ve all seen evidence of either over-optimistic sales forecasts or “sandbagging” to provide slack in a budget. AI is ideally placed to highlight where this might be happening.
There is certainly a cross-over between accuracy and time-saving in relation to AI. Referring to the example above, anomaly detection for journals and sub-ledger values would provide a huge time saving for finance teams, as inaccuracies would be more easily identified.
Imagine revenue planning whereby a host of calculations are done automatically within the AI model’s algorithms, reducing the convoluted process of analysing and applying trends to huge data sets and eliminating some of the human input with driver-based intelligent plans. This time saving would shift focus to the value-added tasks.
How Can Your Finance Team Use AI?
The applications for AI are both broad and far-reaching, thus having the potential to give the finance team the chance to add considerable value to the business.
Anomaly detection is usually an easy start point. Our product is Workday Adaptive Planning, and this has anomaly detection built-in as standard. This means any customer can start to use this almost immediately in their solution with almost no additional effort.
More sophisticated applications would be unique to a business, and the data available. A potentially compelling application for AI is planning for demand. Across all products or services, demand impacts are felt from many different and varied factors, both internally and externally, to the organisation and market. Demand is also at the very core of the success of the business and therefore its value is high and its impact all-encompassing.
When assessing where an AI model could be utilised, we consider questions such as:
- What data does the business have at its disposal that could be of value in planning?
- What other data could be of value, that is not currently in use due to lack of resources and time?
- What factors do we suspect are impacting planning, that we currently are not able to assess properly, due to the sheer volume or complexity of the information?
Obviously, data is at the heart of any solution. The sorts of data that are typically most useful are:
- Historical trends – New product introductions, sales locations, product or service offerings, promotions, buying habits
- Economic & political - Inflation, unemployment, tax rates, currency movements, tariffs
- Social & Environmental – Population and demographics, weather, consumer trends, events
- Customer Interactions – Website and content data, point of sale data, search volumes and other traffic metrics
Our consultants would work with you to identify where AI would add value to your planning process, and how best to shape a project for you.
AI does not apply everywhere. There are examples in planning where there simply isn’t the need for a complex algorithm or neutral network, as the relationship is clear and known. If I need this product, in this location to sell to a customer on this date, I will need to procure this so many weeks prior to order to be able to supply accordingly. This means an easy to explain, formula-based approach is still appropriate in these cases.
In summary, when AI is stripped down to its underpinning fundamentals, it is merely a computer tool that analyses information. It isn’t a pseudo-human wanting to take over the world, or a robot that could replace the humans within FP&A, but a hard-working assistant with a keen eye for finding trends in data.
AI has its place in any modern planning solution and is most certainly here to stay. If you would like to understand more, please contact us to see where it might fit into your processes.