AI is forecast to take over repetitive work processes, but will it lead to more unemployment or create new opportunities for increased efficiencies?
By James Gallaway
People who regularly have to work late into the evening must find grimly ironic the prediction by British economist John Maynard Keynes in 1931 that we would be working three hours a day by the 2030s.
Keynes said constant technological change would lead to more spare time and saw that as a problem for most people “because we have been trained too long to strive and not to enjoy” leisure without a sense of dread.
Whether that fear will be about increased leisure or increased unemployment remains to be seen, but either way Keynes’ prediction may be about to play out. According to the Commonwealth Scientific and Industrial Research Organisation (CSIRO), around 75 per cent of all work done today will be affected in the coming decade by machine learning, automation and artificial intelligence (AI).
The impact on accountants and others working in financial services will almost certainly be profound.
The professions are forecast to embrace new uses for AI in programs such as robotic process automation, which reads documents through optical character recognition and – without human involvement – records and processes the data from daily commercial activities.
“Work practices will be changing more quickly in the future and we have to negotiate two major discourses,” explains Dr Claire Mason, senior social scientist and co-author of a CSIRO study into the future of work in a digital Australia.
“In one, technology is viewed as the agent of change and it is up to workers to get on board. An alternative view is that technology is a tool that humans wield to improve our world. The second is much more empowering.”
Will automation and the speed of change create unemployment? Mason says swathes of mundane and routine tasks that pepper every profession will be taken over by processes involving AI in computing networks and cloud-based systems.
“In the knowledge economy, we see people working in more specialised roles and updating skill sets, so they can use technology to do their jobs more efficiently,” Mason says.
“However, with change occurring more rapidly and frequently, a tertiary degree could be outdated quicker than it’s completed. A lifelong learning model allows workers to tailor skill sets.”
The automation and unemployment connection
Productivity Commission deputy chairman Karen Chester underscored Keynes’ warning of “technological unemployment” at a Committee for Economic Development of Australia (CEDA) meeting in July 2018, when she said that, through all the technological change, we are not seeing “a persistent upward trend in the unemployment rate”, adding, however, that “universities need to improve student employment outcomes – delivering qualifications relevant to labour market needs at a time relevant to the worker’s needs.”
The urgency of this issue is not lost on Professor Michael Davern, professor of accounting and business information systems at the University of Melbourne.
“In education, we’ve got to work out what we are about,” Davern says. “We are not just about content knowledge, but about reasoning skills in the context of that content knowledge. We’ve got to innovate in assessment to better capture adaptability and reasoning skills.”
In Victoria, the Technical and Further Education (TAFE) network now offers advanced Xero training courses in accountancy, with Xero consultants providing industry-relevant approaches to working with software and accounting data. It is the sort of initiative that Laurel Grey, senior digital analyst at audit, tax and consulting firm RSM Australia, cites as part of the general direction of Australia’s vocational education and training sector for quickly adopting technological solutions.
Grey is encouraged by the rate at which Australians adopt technology.
“[The] banking system has been quick to release apps and other ways of processing data, and the Open Banking system promoted by the federal government will only push change more quickly by allowing businesses to access their data.”
Fintech value pack. Includes 10 courses on financial technology (fintech) with a focus on blockchain, bitcoin and other cryptocurrencies, artificial intelligence (AI), and robotics.
Grey’s work at RSM deals mostly with small-to-medium enterprises (SMEs) and she says the majority of her clients are not using cloud technology to its full potential. The situation is worse in rural areas because of data transfer limitations.
“Nevertheless, we have a goal to move all of our clients over to cloud-based storage of data within the next 18 months,” she says.
As more businesses migrate to the cloud, automated software such as QuickBooks and MYOB Essentials, which are interconnected on all devices and synced with live banking data that does not have to be manually entered, will continue to develop, “allowing an open integration with specialised payroll, HR operations, advanced billing and project task management”, Grey predicts.
“Software in this environment takes away a lot of the routine work people [used to have to do] to learn on the job.”
As the potential for even more significant change draws closer, solicitors such as Aitken Lawyers director Walter MacCallum are enthusiastic about the efficiency that artificially intelligent systems can provide by automating routine transactional work.
MacCallum is, however, wary of the risks of cybercrime and has implemented protocols that require voice affirmation from clients before any transactions can proceed.
“It a simple fact that we’re living in a world where cybercrime is an ever-present risk that needs to be minimised,” he says.
Data security remains one of the biggest question marks in AI
Mason and the team at CSIRO Data61 (Australia’s largest data innovation group) identify cybercrime and privacy concerns as key factors that will determine the extent to which new developments are adopted and the full potential of automation realised – or the degree to which it is constrained.
Grey, who is about to take up a board position with the Australian Business Software Industry Association, understands that her clients’ greatest concerns are adequate staff training and security.
“We will make sure vendors are compliant and meeting their security issues,” she assures.
This tension between risk and reward is similar to the issues affecting driverless cars and is analogous to data security problems in cloud computing, according Davern.
“Are self-driving cars completely safe? No, but the ultimate question will be, do they harm less people than human-operated vehicles?”
He says data security is on the same trajectory and there will be data breaches, but the task is to mitigate the risk “by making sure you are less vulnerable than the next guy”.
Is an AI winter coming? Not so fast.
A confluence of forces, according to the CSIRO report, including “device connectivity, data volumes and computing speed, will combine with rapid advances in automated systems and artificial intelligence” to drive change and accelerate its effects in an exponential manner, which is already creating a considerable amount of anxiety.
For their part, Australian CEOs predict growth and increased staff numbers, despite perceived vulnerabilities to cybercrime, and 68 per cent say they are ready to lead their organisations through radical transformation, according to KPMG’s 2018 Global CEO Outlook, titled Growing pains. While most CEOs say they feel “overwhelmed” by time limits to deal with the disruption, 96 per cent see it is an opportunity for growth.
Keynes made his three-hour work day declaration at the beginning of the second machine age, just five years before Alan Turing, British mathematician and computer scientist, first proposed an abstract universal computing machine (soon known simply as the “universal Turing machine”) that was programmable.
AI has been with us for six decades now, and one of the voices warning of the dangers of hyperbole is professor of psychology at New York University Gary Marcus, whose machine learning start-up, Geometric Intelligence, was acquired by Uber in 2016.
"Software in this environment takes away a lot of routine work." Laurel Grey, RSM Australia
In a controversial paper released early in 2018, Marcus warns of an “AI winter, such as the one that devastated the field in the 1970s”, following a critique that claimed AI was too narrow and too superficial to be of practical use.
Marcus says although there are vastly more practical applications for AI now than in the 1970s, currently hyped promises of “a degree of imminent automation that is out of step with reality” could mean executives investing massively in AI may be disappointed.
“Machines cannot in fact do many things that ordinary humans can do in a second, ranging from reliably comprehending the world to understanding sentences,” he says.
Davern maintains AI is not necessarily a threat to the jobs of graduates or experienced business professionals. He sees technology as complementing rather than replacing human abilities.
“For machine learning to get to the point of approaching the work of human consciousness it would have to understand meaning, and it doesn’t,” he says.
“It’s easy enough for a machine learning program to tackle games because the data is readily available and can be fed in repeatedly, but strategic decisions, for example, often have little data history, they are a decision at the point of now that leverages causal understanding rather than just patterns in data.”
Despite their shortcomings, AI and machine learning are still enormously valuable efficiency tools. The extent to which they radically change the way we work will only be confirmed by the work we do with them in the future, and by then another set of technological changes will likely create new levels of surprise and anxiety.
5 ways accountants have mastered AI