We use AI to catch account manipulators.
To date, our opponents have all been human. The time is approaching, however, when we will need to consider what tech-savvy perps might be doing on the other side of the fence. How will our system perform when it competes against machines? Will it be like stumping up for weekend tennis and seeing Novak Djokovic pulling up his socks?
One day this scenario might be real, but when is that day? It depends on how AI evolves in accounting software: Who will build it, how companies will initially use it to augment accounting processes, and the key purposes management will ultimately want AI to serve.
Once AI is well understood and easily used by management, only then will the use of AI for account manipulation become a major concern. For now, it is not, but we think it’s a useful intellectual exercise to examine what the lead-up to that looks like.
If nothing else, so that we can say in a few years (give or take a decade) that “we told you so.”
Since accounting entails a lot of repetitive tasks, and a lot of decisions with simple objectives, one might think that the development of AI for this field should be rapid.
However, before one can develop an AI application, even to manipulate accounts, one must have a software platform that can generate a reliable set of accounts – one must have data. AI applications for accounting will not be stand-alone but rather will be integrated into existing accounting software and processes.
The enormous sunk cost associated with legacy accounting systems means that most companies will want to integrate AI into their existing software. They will look to their existing suppliers for solutions. This process will be slow due to the risks involved.
This task of integration is where most effort will be directed. It is where AI developers will make the most money, because it will be customer-facing. To be sure, there will be new entrants, but the incumbents will have a huge advantage.
It’s likely then that none of the players focussed on developing AI systems for accounting players will be thinking about account manipulation. That is not where the money will be because until there is integration, there cannot be manipulation. Until that point is reached, it will remain far easier to manipulate accounts by fabricating numbers than by spending thousands of hours to develop software.
A recent Wall Street Journal article discusses the AI challenge confronting CFOs. It suggests that CFOs are experimenting with AI-isolated tasks such as “to predict questions that analysts will ask, research competitors’ earnings calls or answer employees’ queries on internal guidelines.” The article highlights concern about reliability, data security and how to use the data effectively.
Most companies are still figuring out how they will use AI in their businesses.
As a general rule, most companies are still figuring out how they will use AI in their businesses. Initially, of course, they will use it in repetitive and rule-based areas such as accounts receivable, accounts payable, audit and compliance. But the accounts department holds all of the data that describes a business.
In the next stage, AI will be used to assess sales trends, monitor inventory usage, assess productivity and costs, and manage procurement.
Ultimately, any management decision made within a company could come under the purview of AI. At some point, therefore, AI will be used not just to manage logistics, but to adjust production plans and even integrate current and cost sales trends with product development.
Recent research from The Hackett Group, a consulting firm, suggests that AI technologies will reduce SG&A expenses by as much as 40% over the next 5-to-7 years. Hackett says that an average company with US$10 billion in annual sales might expect to save more than US$180 million per year in SG&A expense via the adoption of AI accounting solutions.
The simple fact is that companies are going to be too busy implementing AI solutions that generate real economic profit in the coming years, rather than using it to fake their numbers.
Once the low-hanging fruit from AI has been exploited, CFOs and senior managers will be focussed on using AI, which by now will likely be highly customized to their business requirements, to analyse highly complex management decisions.
In most companies, this typically means making investment and divestment decisions, working out how much to pay people, deciding strategies to sell products, and choosing how to fund and account for these decisions.
How a company chooses to fund and report complex decisions has an enormous impact on a company’s reported earnings.
Traditionally, the CFO sits at the nerve centre of a company’s data, its manual processes and key decisions. At some future point, AI will sit at the nerve centre.
Traditionally, the CFO sits at the nerve centre of a company’s data, its manual processes and key decisions. At some future point, AI will sit at the nerve centre. In principle, the CEO will be able to make decisions that once required the CFO and a team of minions to answer. It may well be the case that account manipulation will become easier in such a world. The fewer eyes and ears associated with decision making, the lower the likelihood accounting manipulation will be detected.
In other words, at some point, every CEO will have a dial on her desk which simply says “accounting aggressiveness.” Evidence suggests that about one in ten public companies will have that dial turned as far to the right as possible. Some will have a red button, covered with Perspex that simply says “make it up.”
Initially, AI software will likely take account reporting to the legal limit due to the constraints of the software systems they employ. But, once accounting fraud detection software such as ours is in wide use, this will not be sufficient. Account manipulators will need software that reverse engineers our software to determine how it scores manipulation risk. The accounts will then be tweaked to give the best possible account manipulation score.
It will be an arms race. Technology always is. But, in this case, it is a race the good guys can win because a company’s accounts must always add up. No matter what account manipulators do, their numbers will give them away.