While banks are positioning themselves to automate the majority of their workforce over the next few years, Australia’s biggest lender is making a very public and embarrassing U-turn. The Commonwealth Bank of Australia (CBA) is being forced to offer 45 staff their jobs back after it was ordered to admit that it lied about making them redundant due to the introduction of artificial intelligence.

The case dates back to when CBA called dozens of long-term staff into meetings to tell them their positions were being cut. The bank had recently launched a new AI “voice bot” chatbot, which it claims reduced incoming call volume by some 2,000 per week. According to the bank, that meant the same number of customers were being serviced with far fewer human staff.

But the news was a shock to the staff in question. Some had been working for the bank for several decades. They quickly questioned the basis on which the bank said it had determined that so many jobs had become redundant. Staff say that the number of incoming calls had not been falling at the time of the layoffs, as the bank claimed. Instead, they suggest that call volumes were actually rising. In fact, employees claim that managers had been drafted in to field calls and that staff were being asked to work overtime to cope with the volume.

The union representing the staff upped the ante when it went to a fair work tribunal, after alleging that the bank had misled both staff and the public about its artificial intelligence technology. The Finance Sector Union (FSU) says the bank failed to adequately explain how the jobs had become redundant and is seeking remedies to support the affected workers.

In fact, the bank is also being accused of using the “excuse” that jobs were being cut because of a new AI voice bot to disguise a decision to move some of the roles offshore to India. It points to the fact that CBA was hiring staff in India at the same time as saying it was making jobs redundant at home in Australia. By alleging that AI technology had become a more efficient solution to answering incoming calls, the union says CBA has created the perception that the chatbot was a smokescreen to outsource jobs.

Appearing before a fair work tribunal, CBA was forced to admit that while it believed it had used its “voice bot” to reduce incoming calls, that was not the case. The bank accepted evidence that there was an unexpected and sustained rise in call volumes that began around the same time the workers were given their termination notices. The rise in call volumes has continued for several months, meaning that CBA’s explanation for cutting those jobs was at best incomplete.

“This error meant the roles were not redundant,” the bank said in a statement before the tribunal. It is a major backdown by CBA, which has since apologized to staff for its mistake. Affected employees are now being offered the chance to return to their previous role, to take another role in the organization, or to take 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 spokesperson said.

The union has called the ruling a “massive win” for its members but says that much of the damage to workers had already been done. Workers were subject to weeks of uncertainty and stress about their livelihoods as the case ran its course. The union also highlighted the impact on workers who had been told to leave the business when job cuts were first announced. The shock that they had provided so many years of loyal service, but now suddenly faced the prospect of not being able to meet their bills, had a long-lasting impact on some.

The union believes the episode should act as a warning to other employers that it is not acceptable to rush the introduction of new technology and make dramatic changes to the workforce without taking into account the human impact. The CBA saga plays out as the bank’s AI ambitions continue unabated. Last week, the bank announced that it was to partner with OpenAI to develop more advanced generative AI tools for banking. The new AI capabilities will be used to help improve scam and fraud detection, to make payments safer, and to deliver more personalized customer experiences.

“We are excited to be working with OpenAI and reaffirms our commitment to making these tools work for Australians, embedding the responsible use of AI,” CBA said in a statement.

Financial services firms are expected to shed a significant number of jobs in coming years as AI and automation move in. In the United States alone, banks may cut as many as 200,000 jobs in the next three to five years as roles in back office, middle office, and operations positions are automated, according to a report from Bloomberg Intelligence.

But the CBA case provides a powerful example of how mistakes during the transition to AI can cause reputational damage, and damage the trust of both staff and customers. For now, CBA workers who were told their positions were being cut will have to decide whether to return to a job that they thought had been eliminated. Many are expected to still walk away, says the union, which points out that the episode has undermined trust in management. “The damage has already been done,” it said.

The union confirmed that while the case against CBA over these 45 jobs has been settled, it is currently also pursuing a second case over the bank’s use of AI. This case is being brought to the Fair Work Commission and concerns the bank’s obligations to consult over the rollout of the technology. The case could yet have implications for CBA’s AI ambitions.