Global investment in information technology has doubled in the past several years. Despite public commitments and laudable efforts, governments and organizations continue to deal with recurring disasters in their software endeavors. Michigan’s MiDAS unemployment system and Australia’s Centrelink “Robodebt” welfare system are instructive cautionary tales. In doing so, they pull back the curtain on the risks inherent in automated decision-making. This was how the Canadian federal government introduced their Phoenix payroll catastrophe back in April 2016. This system is one of the largest IT failures to linger on and continue to affect more than a hundred thousand dirty public servants today.
Global IT spending has more than tripled since 2005, reaching a staggering $5.6 trillion. Yet, by many measures, the success rates of software projects improve little, if at all. A report from 2024 revealed that 80 percent of organizations acknowledge that inadequate or outdated technology stifles progress and innovation. The very organizations doing the incredible work are under enormous strain. It isn’t hard to see where failures result, not from technical incompetence, but from incompetent management and unrealistic expectation.
Case Studies of Automated Decision-Making Failures
The MiDAS unemployment automated decision-making system in Michigan illustrates the dangers of such automated decision-making. Even further, Australia’s Centrelink “Robodebt” initiative echoes these terrible outcomes. These systems are intended to increase efficiencies in welfare administration. One after another, they’ve sparked lawsuit after lawsuit — and public outcry — because they miscompute claims.
Welfare recipients were calculated debts by a failed algorithm. Consequently, hundreds of thousands of people were incorrectly targeted to repay funds. Particularly, the fallout from such mistakes shattered public confidence and had serious legal repercussions for the Australian federal government.
The Phoenix Payroll System: A Cautionary Tale
The Phoenix payroll system has since become a cautionary tale within Canada’s governing institutions. When it was launched in 2016, project managers overlooked the lessons learned from the original payroll system failure in 1995. This oversight turned out to be costly, as three out of four of the 430,000 current and former federal employees had payroll issues.
As of March 2025, more than 349,000 errors pertaining to the Phoenix system remain unresolved. Even more alarmingly, fully 53 percent of these issues have been pending for more than a year. The incident exemplifies the impact of misguided management practices and unattainable project expectations. Like many of their peers, the confident Phoenix team knew that they could design a better, more modernized payroll solution. They endeavored to tailor one off-the-shelf package to tackle the 80,000 pay classifications in 105 collective agreements.
The team continued on with the Phoenix project, undeterred by the advice. They did it on a budget that was under 60 percent of the vendor’s initial cost estimate. This decision serves as an opportune reminder of the perilous dangers embedded in underfunded, overpromised projects.
“Anyone can make a mistake, but only an idiot persists in his error.” – Cicero
Addressing IT Failures: A Path Forward
So far the Canadian government’s response has acknowledged the seriousness of the issues created by the Phoenix payroll system. They’re committed to resolving their backlog of mistakes by June 2026. Software failures usually result from human imagination and poorly managed risks, and this commitment signals that we understand that reality. They certainly aren’t due only to technical malfeasance.
State fiscal reports have been shedding light on this trend in recent months. As recently as fiscal year 2023-2024, a third of all civil servants were subject to paycheck errors because of the continued failures of the Phoenix pay system. This ongoing struggle serves as a constant reminder of the necessity for diligent oversight and accountability in all IT endeavors.
Experts like Bent Flyvbjerg from the University of Oxford’s Saïd Business School emphasize that IT projects are among the riskiest in terms of cost. Now is the time for organizations to invest heavily—but realize that they need to manage their expectations in line with reality.
“To Engineer Is Human: The Role of Failure in Successful Design” – Henry Petroski