As banks prepare to automate substantial sections of their workforce, Australia’s biggest bank has made an embarrassing U-turn. Australia and New Zealand Banking Group Ltd. (ANZ) has been forced to rehire 45 workers after initially telling them they had been made redundant by artificial intelligence. In a decision that was forced by a dispute at the Fair Work Tribunal, the Commonwealth Bank of Australia (CBA) was found to have misled both its employees and the public regarding the capabilities of its new chatbot technology.

The case dates back several months to when the bank began contacting dozens of long-standing workers to inform them their jobs were no longer necessary. In its explanation, CBA stated its recently introduced AI-powered “voice bot” had decreased incoming calls by about 2,000 per week, meaning there was no longer a requirement for so many human staff. Some of those who were dismissed had worked for the bank for up to 30 years, making the news even more difficult for those involved to take.

However, these employees were not going to accept that explanation. Instead of a reduction in call volumes, staff claimed that these had actually increased at the time the redundancies took place. In fact, management was reportedly in the process of mobilizing resources to meet the demand, including temporarily reassigning managers to take calls and offering overtime to existing workers.

The union in question then took up the case with the Fair Work Tribunal. The Finance Sector Union (FSU) argued CBA had not adequately justified the roles as redundant. It also said there was some evidence that the bank may have been using the chatbot as a pretext for a move to offshore some roles to India, as new hires were also taking place in that country at the same time.

In its evidence to the tribunal, CBA then accepted a crucial point. Bank representatives stated that they had not considered that a sustained increase in call volumes had coincided with the redundancy decision. A sharp rise in calls, which continued over a period of months, directly contradicted its earlier reasoning for the layoffs. “This error meant the roles were not redundant,” CBA conceded in the tribunal.

As a result, the bank was forced to back down and issue an apology to affected staff. CBA now says that these 45 employees will be given the option to return to their old jobs, be redeployed into new positions in the bank, or given an exit package. “We have apologized to the employees concerned and acknowledge we should have been more thorough in our assessment of the roles required,” a bank spokesperson told Bloomberg.

The union has called the victory a “massive win” for its members, but it also argued that the damage had already been caused. Many of the staff involved had to live with weeks of anxiety about their future, some of them suddenly confronted with the possibility of not being able to pay their rent or their bills. The FSU also argued the entire case showed the risks that arise from employers rushing to deploy AI without thought to its impact on staff.

CBA’s move to U-turn the layoffs does not appear to have slowed its march towards a greater reliance on AI. Just last week, CBA announced a new partnership with AI company OpenAI. Under this deal, the two companies will work together to develop generative AI models that can be used to strengthen the bank’s capabilities in detecting scams, preventing fraud, and providing more tailored services for customers. CBA says that the new tie-up is about “investing in its people and embedding the responsible use of AI”, but the announcement has been met with a healthy degree of skepticism from staff members who were misled over the chatbot in the first place.

As this latest controversy demonstrates, the use of artificial intelligence is transforming the way banks and financial institutions do business. Bloomberg Intelligence estimates that global banks alone could be at risk of shedding 200,000 jobs over the next three to five years. As AI and automation take over back office, middle office, and operational roles, banks see this as an opportunity to both cut costs and drive greater efficiencies. In doing so, CBA may have made a series of mistakes that will have reputational ramifications, as well as a cost that will not be fully accounted for in its financial statements: the loss of trust with employees, and by extension, customers.

For now, the 45 employees at the center of the dispute must make a choice about whether they want to return to the jobs they were told did not exist. The FSU said it expected a large number would still choose to walk away, arguing the incident has undermined trust in management beyond repair. “The damage has already been done,” the union said. For now, it is a reminder that for workers, change can come quickly, and no one is completely safe from the disruption that automation can cause.

Union has confirmed that, although this battle is over, another is on the way, as it is also pursuing a case at the Fair Work Commission regarding CBA’s consultation obligations when it comes to its broader use of AI. Whether that case will help to slow or stop the bank’s progress towards automation is yet to be seen, but one thing is already clear: if the march to AI-enabled banking is as disruptive and widespread as many now expect, then its path will likely be a great deal messier than banks or executives were expecting.