Let’s cut to the chase. Of the 702 occupations that University of Oxford academics Carl Benedikt Frey and Michael A. Osborne recently examined for susceptibility to computerisation, accountants and auditors are judged to have a 0.94 probability of being at risk.
This puts CAs and CPAs on par with tax preparers, data entry keyers, loan officers, bookkeepers, accounting and auditing clerks, cashiers and budget analysts as the most endangered occupations in the near future.
In contrast, the jobs that are at low risk of being replaced by automation include management analysts, compliance officers, marketing managers, sales managers and CEOs (see table below).
Good-bye, Real-Life Accountant?
Sources: The Future of Employment: How Susceptible Are Jobs to Computerisation? and CFO Innovation
No one can say they didn’t see this coming. In a world where driverless cars are coming on the road, the tasks of posting, reconciling and reporting financial results would seem to be particularly susceptible to automation. Come to think of it, when was the last time you met a bookkeeper in your organization?
The job of the accountant and auditor is more complex, of course. Some cognitive skills, not just routine processing and manual dexterity, are needed to examine, analyze, and interpret accounting records to prepare financial statements, to audit and evaluate statements prepared by others, and to install or advise on systems of recording costs or other financial and budgetary data.
But even such skills, reckon Frey and Osborne, are becoming susceptible to computerisation. In "The Future of Employment: How Susceptible Are Jobs to Computerisation?", they argue that advances over the next decades in mobile robotics and machine learning technologies such as data mining, computational statistics and other sub-fields of artificial intelligence will produce algorithms that allow cognitive tasks to be automated.
Already, sophisticated algorithms are taking over a number of tasks performed by paralegals and contract and patent lawyers, they point out. For example, law firms now utilize computers to scan huge volumes of legal briefs and precedents to assist in pre-trial research.
Artificial intelligence algorithms, the academics add, are now “able to process a greater number of financial announcements, press releases, and other information than any human trader, and then act faster upon them.”
“Although the extent of these developments remains to be seen,” concede Frey and Osborne, “estimates by [the McKinsey Global Institute in 2013] suggest that sophisticated algorithms could substitute for approximately 140 million full-time knowledge workers worldwide.”
How did the academics come to assign such a precise probability – 0.94 (1 = computerisable; 0 = not computerisable) – to the job of accountant and auditor? The susceptibility of these posts to computerisation was subjectively estimated at “1” by Frey, Osborne and a group of machine learning researchers.
Seventy of the 702 occupations, including accountant and auditor, were actually subjectively hand-labelled, based on open-ended descriptions of tasks specific to each job, as provided (and regularly updated) by the US Department of Labor through its O*NET online service.
The hand-labelling was made by answering the question: Can the tasks of this job be sufficiently specified, conditional of the ability of big data, to be performed by state of the art computer-controlled equipment?
“We drew upon a workshop held at the Oxford University Engineering Sciences Department, examining the automatability of a wide range of tasks,” the academics explain. “Our label assignments were based on eyeballing the O∗NET tasks and job description of each occupation.”
The susceptibility to computerisation of all 702 jobs, including accountant and auditor, were then measured using an algorithm developed to assess probability based on nine variables: persuasion, negotiation, social perceptiveness, originality, assisting and caring for others, fine arts, manual dexterity, finger dexterity, and cramped workspace.
Three Gaussian process classifiers were tested and the results compared with the hand-labelled probabilities assigned to the 70 jobs. The classifier whose results most closely matched the hand-labelled probabilities was selected – this was the exponential quadratic model, whose AUC (area under the receiver operating curve) was equal to 0.894 (1.0 is the perfect classifier).
In general, said the authors, the algorithm found that the following variables exhibit “relatively high values in the low risk category” – meaning that they are associated with low susceptibility to computerisation:
assisting and caring for others
In contrast, the following variables showed low values in the low risk category, meaning they are associated with high susceptibility to computerisation:
cramped work space
Frey and Osborne concluded that “generalist occupations requiring knowledge of human heuristics, and specialist occupations involving the development of novel ideas and artifacts, are the least susceptible to computerisation.”
They add: “Most management, business, and finance occupations, which are intensive in generalist tasks requiring social intelligence, are largely confined to the low risk category. The same is true of most occupations in education, healthcare, as well as arts and media jobs.”
What About the CFO?
Frey and Osborne include “finance occupations” in the category of jobs that carry a low risk of being replaced by computers, such as financial analysts, statisticians and actuaries.
But not accountants and auditors. Apparently the O*Net descriptors of their core tasks, according to the algorithm, correlate highly with the variables that denote high susceptibility to computerisation.
For example, one core task of an accountant is to “establish tables of accounts and assign entries to proper accounts,” which is an activity that relates to manual dexterity and not so much to, say, negotiation, persuasion or originality.
On the other hand, O*Net describes one core task of an auditor as collecting and analyzing data to detect deficient controls, duplicated effort, extravagance, fraud, or non-compliance with laws, regulations, and management policies.
This is strikingly similar to the automation of the task of scanning large volumes of legal briefs and precedents by law firms, and of financial announcements, press releases and other information by market trading firms.
As Frey and Osborne write: “Fraud detection is a task that requires both impartial decision making and the ability to detect trends in big data. As such, this task is now almost completely automated.”
What about the CFO? O*Net includes this job title under the category “Treasurers and Controllers,” along with Finance Vice President, Director of Finance and Finance Manager. But Frey and Osborne chose not to include these jobs in their study for technical reasons.
Still, the core tasks O*Net ascribes to these occupations seem associated with the low-susceptibility-to-computerisation variables rather than to manual dexterity. For example, one of the CFO’s core task is to “advise management on short-term and long-term financial objectives, policies, and actions,” which has elements of originality, negotiation, persuasion and social perceptiveness.
Arguably, with the expansion of the finance function’s remit to include analytics, strategy and business partnering, the job of the CFO is taking on some aspects of the tasks of the CEO. And the occupation of chief executives, according to Frey and Osborne, has a very low probability, at 0.015, of being replaced by a computer.
Are accountants and auditors doomed? Not necessarily. “We do not capture any within-occupation variation resulting from the computerisation of tasks that simply free up time for human labor to perform other tasks,” the academics acknowledge.
A case can be made that the automation of the core tasks of accountants and auditors could help them move on to higher value activities that, currently, are categorized as merely supplemental, not core, by O*Net, such as:
Advise in areas such as compensation, employee health care benefits, the design of accounting or data processing systems, or long-range tax or estate plans
Develop, maintain, and analyze budgets, preparing periodic reports that compare budgeted costs to actual costs
Analyze business operations, trends, costs, revenues, financial commitments, and obligations, to project future revenues and expenses or to provide advice
Advise about issues such as resource utilization, tax strategies, and the assumptions underlying budget forecasts
Survey operations to ascertain accounting needs and to recommend, develop, or maintain solutions to business and financial problems
The case is arguably weaker for the auditor, based on the core and supplemental tasks for this occupation listed by O*Net. But the assurance and accountability aspects of auditing can possibly save the job.
A trusted third party would be needed to make sure the algorithms used in the automation of auditing processes are legitimate and lead to conclusions that a human auditor would have reached. Regulators may also require that a human or at least a corporate entity should be there to be accountable for flawed certifications.
New expectations of businesses can also reinvent the job of the accountant and auditor. Integrated reporting, which in theory would require connecting the dots between financial results and environmental, social and governance activities, will involve subjective and qualitative judgments.
This may be the most potent route for accountants and auditors to remain relevant in the coming decades. Let algorithms take over the routine processes and use the outcomes to judge, certify, advise, persuade, assist and transform, including in the areas of climate change, risk management and regulatory compliance.
There is still hope for the profession.
About the Author
Cesar Bacani is Editor-in-Chief of CFO Innovation.