As Australian businesses delve into the advantages of generative AI—such as enhanced productivity and improved employee experiences—concerns about potential risks are emerging, especially among workers unfamiliar with these challenges. A recent survey by Google, conducted with IPSOS, revealed that nearly 50% of Australians are utilising generative AI, with about 75% of these users applying the technology at work. Furthermore, a study from HR platform Workday discovered that approximately 65% of employees confirmed that their workplaces had integrated AI technology.
Despite its benefits, using generative AI for routine tasks like sending emails can lead to serious setbacks. Michael, the director of End of Lease Cleaning Melbourne, experienced a significant financial loss due to an error stemming from an AI-generated email. His team aimed to streamline their response time from five hours to one by using generative AI to create tailored cleaning quotes based on customer requests. However, the tool mistakenly referenced services, leading Michael’s business to provide expensive cleaning options at discounted prices.
After a series of costly mistakes, including a $2000 loss for a particularly challenging cleaning job, Michael soon realised the risks of relying too heavily on AI. His oversight of an email, which contained several inaccuracies, resulted in a lost contract, confirming that such errors could jeopardise business prospects. Similar instances of miscommunication have been reported by others, where AI addressed clients incorrectly, causing misunderstandings and damaging professional relationships.
Legal experts Robyn Chatwood and Michael Park warned that AI tools sometimes “hallucinate,” producing inaccurate information that can muddy professional correspondence. Additionally, there are concerns regarding possible copyright infringements and breaches of confidentiality due to AI’s generative capabilities. If mistakes arise, the onus ultimately lies with the worker who implements them, as Chatwood noted, emphasising the need for diligence in verifying AI outputs.
To mitigate such risks, Park advises workers to adhere strictly to their employer’s AI policies, suggesting that even if AI seems to offer an advantage in efficiency, it is crucial to remain compliant. Murray’s team has since decided against using AI for business emails, reverting to their previous slower response times to avoid further errors.
Michael has learned that thorough proofreading—two to three times—of AI-generated content is essential before sending it out. Ultimately, while the implementation of generative AI has the potential to enhance workplace efficiency, users must remain vigilant about the accompanying risks.