In July 2017, I presented at the 2017 AICPA FP&A Conference in Las Vegas over AI...
Artificial Intelligence and FP&A
In a recent ACCA survey about the future of the accountancy professional, digital is listed as one of the 7 quotients for success. ACCA defines digital as “the awareness and application of existing and emerging digital technologies, capabilities, practices, strategies and culture”. One of the more talked about area in technology is artificial intelligence. This article attempts to explore whether artificial intelligence can replace people in the FP&A arena.
Artificial intelligence involves using computer systems to perform tasks that normally require human intelligence. Artificial intelligence is not the same as automation. Intelligence involves the ability to infer from a limited amount of data/information in making decisions. Machine learning is required to engage AI in an effective manner. AI is like a child – a set of rules are needed for the system(s) to learn and develop the appropriate analytical processes. AI has yet to evolve to be creative and deliver new solutions to issues identified through the analysis.
Most organisations use technology to assimilate Big Data and produce insightful analysis. Organisations collect data because they want to get a better insight into their market, customers, products, etc. Supermarket giants make use of POS data to assess buying habits and determine the required stock level. Businesses often already have an idea of the insight and needed data to support their hypothesis – supermarkets already know barbecue products are particularly popular during the summer months. The use of technology to process, categorize and analysis Big Data in this scenario to deliver insight can be seen as self-justifying. I also recall a time when I was engaged to perform some analysis on product profitability. The sponsor of this already had an idea about whether the products were profitable. My exercise was to prove (or disprove) the sponsor’s hypothesis. If AI were used to conduct this analysis, it would yield the same outcome – it was not really generating an insight but more about proving a hypothesis.
The use of AI for analytics is to take advantage of the powerful processing power and remove the human factor from the equation. This means human bias is “eliminated” when producing analysis. However, people are not totally removed from the process: it is people who wrote the rules/codes and it is people who make the final choice and execute the chosen solution. For example, Power BI users write DAX formulae in their BI model to enable more complex processing of large volumes of data. These DAX formulae are defined by humans, not a machine. Let’s imagine, the user wrote the DAX formula for calculating profit with the wrong field, the machine will not be able to tell it is an error.
Another aspect of using AI in data analysis is the development of new solutions. AI follows a set of decision tree rules. When AI encounters a situation, it has not encountered before, it may not always be able to generate an appropriate solution and is likely to defer to human. Machine learning can only improve when a human is “reintroduced” into the equation through giving feedback which enables the machine to gain “experience”. This feedback loop is not always present. Therefore, machine learning can be limited. For example, in some visualisation tools, users can ask questions to interrogate data. The creates an interactive feel but relies on the models recognising keywords. If the keywords are not fed back into the machine’s vocabulary base, the interaction is lost.
As technology evolves, FP&A should evolve to harness the opportunities that technology presents. The increase in computer processing power means that more data can be processed and analysed. The resulting insights can stimulate discussions and thinking. This enables decision-makers to consider areas that may have otherwise been missed. AI can be seen as a disruptor. Disruption is known to have stimulated improvements. AI in its current state takes away the mundane laborious data collection and processing tasks. This gives FP&A more time to look at value add activities such as business relationship building, value add advisory work and invest time in process improvement. FP&A is likely to evolve from producing insights to actively engaging their business partners to develop the business. With automation, FP&A has already saved time in processing data. With artificial intelligence, FP&A should look at being involved in setting the rules of engagement.
A final thought – will we really have more free time when true AI arrives?
Any views or opinions expressed are solely those of the author and do not necessarily represent those of The Warranty Group. The Warranty Group is part of Assurant.
The article was first published in Unit 4 Prevero Blog