More than half of Australian workers distrust AI data

New research from Salesforce has revealed that more than half (52%) of office workers in Australia find it challenging to obtain the desired outcomes from artificial intelligence (AI). The research, presented today in Melbourne, also highlighted that 52% of workers do not trust the data used to train AI systems. These insights were part of the AI Trust Quotient study.

The research involved a survey of 1,038 full-time office workers in Australia and nearly 6,000 globally. A significant finding was that a substantial 89% of Australian respondents do not currently trust AI to function without human oversight. This trust deficit is proving to be a barrier to the broader adoption of AI technologies; 62% of those who do not trust AI indicated hesitance in utilising it.

For businesses to fully capitalise on AI, the study suggested that reliable data inputs and consistent human oversight are vital. Salesforces's AI Trust Quotient highlights several critical findings related to trust and adoption issues concerning AI in Australian workplaces.

One key point is the role of unreliable data in perpetuating the trust gap. Over half of the respondents find it challenging to get what they want from AI at present, and a similar percentage express distrust towards the data used to train AI systems. Fear that humans will lose control of AI also remains prevalent, with 62% of the surveyed office workers sharing this concern.

Another important aspect is the necessity for businesses to prioritise good data management. According to the research, 79% of those who do not trust AI attribute their mistrust to a lack of useful information. Public data that is out of date and incomplete customer or company data were also cited as barriers by 71% and 66% of respondents, respectively. Moreover, 88% of office workers stated that the accuracy of the data used by AI tools is crucial, while 87% stressed the importance of data security. A slightly smaller majority, 84%, said that comprehensive and relevant data would help build trust in AI systems.

For AI to gain wider acceptance, it must consistently deliver accurate outputs; 73% of office workers indicated that consistent accuracy is essential for building trust in AI, a higher threshold compared to the global average of 68%. Hence, Australian businesses need to address data silos and ensure that their AI systems are well-integrated with the company's own data.

As AI technology becomes more sophisticated, maintaining human oversight will remain essential to harness its full potential. The survey outlined that 89% of office workers do not trust AI to maintain data security autonomously, but almost half (49%) trust AI to keep data safe when humans are involved alongside the technology. Skill enhancement and training opportunities (67%) and greater accessibility and inclusivity (62%) were identified as human-led factors that could drive more trust in AI.

Commenting on the research findings, Rowena Westphalen, Senior Vice President of Innovation, AI & Customer Advisory APAC at Salesforce, noted that trust in AI is crucial for employee acceptance. She emphasised the value Australian workers place on human oversight in AI operations, highlighting the combination of machine efficiency with human judgement and creativity as a pathway to a more productive future.