Accountants Must Pioneer AI Solutions
South African Institute of Professional Accountants
16 January 2019
There’s been much talk in the press about the robotisation of accounting and how artificial intelligence (AI) will inevitably make professional accountants obsolete. According to Faith Ngwenya, Technical and Standards Executive at the South African Institute of Professional Accountants (SAIPA), it’s not that simple. “Professional accounting organisations like IFAC and SAIPA are closely monitoring the impact of AI-driven systems on the industry,” she says. “While some bookkeeping and reporting functions may require minimal human intervention in the future, the professional accountant who is willing to expand their horizons will be more in demand than ever.”
AIs in training
To date, there are no commercially available general AIs, that is, ones that can learn to do anything a human can. Rather, today’s AIs must be trained to perform an individual task by analysing historical data related to that function. They also have to be given statistical boundaries so they know whether a given piece of information is relevant to completing the task or not. And they must be configured to recognise when they have reached the desired goal. Once set up, an AI continually executes the task by trial and error in a training environment until it achieves an acceptable rate of success. Only then can it be deployed to a live business environment. If the task is unique and limited data is available, system and data preparation could take several months.
The accountant’s role in financial AI projects
If the provided training data is insufficient or not strictly relevant to the task, or the configuration parameters are inappropriate, an AI may never reach its goal. Worse still, an incorrectly trained AI might complete the task without fail, but with compounding inaccuracies that go unnoticed until it’s too late. For financial functions, this scenario gives rise to an opportunity cost, risk to the business and concerns about protecting the public interest. “The correct training of an AI to perform even basic accounting work clearly falls within the professional accountant’s purview of developing, implementing and auditing financial systems,” says Ngwenya. “They must therefore be prepared to study and understand at least the rudimentary principles of AI because the accountability for its results falls on them. AIs don’t sign off on internal audits or financial reports.”
A new opportunity for accountants
How AI can be applied to complex business functions and problems is still largely a matter of human ingenuity. Apart from guaranteeing that an AI’s work is accurate, accountants must also ensure that the way the technology is used adds value to a company’s financial processes. Being exposed to the full body of business data and requirements, they are in a prime position to drive business automation in a direction that best serves their customers or employers. Private practitioners may partner with AI specialists to develop useful business solutions or even hire employees with these skills. For example, one company has launched an AI that can rapidly detect embedded leases in contracts, an expensive and labour-intensive requirement of IFRS 16.
Says Ngwenya: “Instead of worrying about AI, professional accountants should pioneer the way forward by offering AI assurance services and innovating the future of automated accounting. This will only enhance their standing as trusted business advisors.”