The world’s banks are gearing up for a drive to automate major swathes of their workforce, but one of Australia’s largest lenders is already reeling from an ugly PR faux pas. Commonwealth Bank of Australia (CBA) will rehire 45 employees after sacking them because their roles had become redundant due to artificial intelligence.

The backtrack follows a complaint by the Finance Sector Union (FSU), which argued the bank had misled both staff and the public about the extent to which it uses chatbot technology. The dispute started when the bank informed dozens of employees, some of whom had served the bank for decades, that their jobs were no longer required.

The lender later said a newly launched artificial intelligence (AI) powered “voice bot” was receiving 2,000 fewer calls per week, meaning it no longer required as many human operators to answer them. In an apparent attempt to smooth over the change, the bank argued that with their old jobs gone, the workers were entitled to take redundancy payments and would now be offered new roles.

However, the affected workers rejected the explanation. They also accused the bank of being economical with the truth about its own technology. Staff told the union the reality was that, far from seeing a reduction in call volumes, they were actually rising when it sacked employees. In fact, managers had been so overwhelmed with demand that they were temporarily redirecting in-house staff to field customer calls, and offering overtime to those who could not leave their desks.

The Fair Work Tribunal later heard from the bank that it had been caught out by a persistent rise in call volumes that coincided with the redundancies, one that it did not expect to last. The spike, which continued for several months, would have directly contradicted its original rationale for the layoffs. “This error meant the roles were not redundant,” the bank acknowledged during its tribunal submission.

The admission forced CBA to backtrack. The bank apologized to the affected employees and said the 45 staff members in question would be offered the chance to return to their old roles or apply for other jobs in the bank, or receive a redundancy payment. “We have apologized to the employees concerned and acknowledge we should have been more thorough in our assessment of the roles required,” the bank said in a statement to Bloomberg.

The FSU has hailed the decision as a “massive win” for its members, but also pointed to the significant damage to staff confidence that had already been done. The uncertainty of suddenly not having a job had hung over the employees for weeks, leaving some to wonder how they would be able to pay their bills. The case has been seen as a clear warning that pushing ahead with AI plans at the expense of workers can backfire.

There is no sign that the bot fiasco has dampened the bank’s AI plans, however. The day after the case, CBA unveiled a new collaboration with OpenAI to develop more advanced generative AI models. The focus of the partnership will be on improved scam detection and fraud prevention, and also on creating more personalized service for customers. CBA has maintained that the deal is about investing in staff and developing a greater focus on responsible use of AI in the bank.

This incident is a small piece of a much bigger picture, however. Financial institutions worldwide are already bracing for a major shake-up. Banks globally are set to cut 200,000 jobs in the next three to five years as AI and automation replace office, middle office, and operations staff, according to Bloomberg Intelligence. While AI can be used to slash costs and increase efficiency, the case of Commonwealth Bank has raised questions over the reputational risks of bungled automation drives.