One of the largest banks in the world is dealing with an embarrassing setback after announcing thousands of workers would be made redundant through the use of artificial intelligence. Australia’s Commonwealth Bank of Australia (CBA) is now being forced to rehire 45 employees whom the bank previously said were no longer needed due to a new “voice bot.” The decision was made after a finance union pushed back on the bank’s claims that the bot was the reason the employees lost their jobs, and accused the bank of misleading the public and the employees about the impact of its new AI tool.

The CBA’s reversal comes after the Finance Sector Union (FSU) submitted a case to the Fair Work Commission, the country’s tribunal and dispute resolution service. The union had previously accused the bank of not providing enough context as to how the 45 roles were made redundant. A hearing later revealed that incoming call volumes to the bank had not in fact decreased, but had increased around the time that the layoffs were announced. This came as a surprise, as bank management was understood to be trying to cope with higher volumes, moving managers to work in call centers and offering overtime to deal with higher-than-usual demand.

The bank later stated in a tribunal that it had not taken into account a volume change when it decided to make the employees redundant. “This error meant the roles were not redundant,” the bank said. “The Commission accepts CBA’s evidence and agrees it made an error in its calculation of calls per staff member,” the CBA added in a statement provided to Bloomberg. The error led the bank to back down from the cuts. In a statement, the bank apologized to affected staff. It also stated that the 45 staff would be offered the choice of returning to their previous roles, applying for a different role at the bank, or accepting an exit package.

The FSU, which has long challenged the bank’s move to bring in AI, celebrated the news as a “massive win” for the staff. The union, which was also involved in a case that helped to drop child safe account software announced by the CBA in August, stated that despite the outcome of this case, the damage had been done. Staff had spent weeks unsure of their livelihoods, and some had to deal with the immediate shock of not knowing if they would be able to pay their mortgages and bills.

The episode has been cited as an example of how employers can push through artificial intelligence solutions without considering the human consequences of such decisions. The bank has stated it is exploring artificial intelligence, both in this recent partnership with OpenAI as well as the deployment of generative AI tools. It has also added that it is implementing new technology to better fight scams and deliver better services to customers. But trust has been eroded, both by the use of the AIs and the bot that was the cause of this case.

The debacle is the latest in a line of them from the bank, with a previous incident where it dropped children-safe bank accounts software after the FSU submitted a similar case. In the same statement, the union said this case has taught banks that consulting with staff over AI tools was not “optional, but a legal requirement.” But the bank will likely continue to bring AI in despite this mistake, as 200,000 jobs are at risk of being cut globally as banks automate middle-office and back-office operations with technology.