Study Finds Most Australian Workers Spot Workplace AI Errors but Fear Speaking Up

A fresh study into artificial intelligence in the workplace has revealed a growing trust divide across Australian organisations. According to research released on 8 October by enterprise software vendor Appian, 60% of Australian workers have caught an AI tool making an error while on the job.

Despite mistakes being common, the findings show a troubling cultural barrier inside businesses: more than one in five employees reported that their workplace discourages staff from questioning or challenging AI-generated advice and decisions. With staff noticing blunders but feeling silenced, only 14% of workers say they fully trust the AI outputs generated within their workplace.

The research underscores that these slip-ups carry real business costs. Nearly half of the respondents who spotted errors said the most serious incident went beyond minor annoyance, creating consequences for compliance, finances or company reputation. Specifically, 20% said the worst failure triggered a customer service problem, while 13% reported direct operational disruptions.

When employees lose confidence in technology, productivity gains quickly evaporate. Staff either spend extra hours second-guessing and verifying every sentence and figure generated by the software, or they let questionable outputs pass through to customers because they worry about challenging automated systems.

What this means for Australian businesses

For small and medium enterprises rolling out generative AI and automation tools, having great software is only half the battle. Company culture and governance dictate whether those tools save time or create fresh liabilities.

• Empower staff to push back: Make it clear to your team that human judgement always trumps automated output. Staff should never feel penalised for questioning a summary, formula or draft prepared by an algorithm.
• Establish clear verification checkpoints: Avoid handing unchecked autonomy to customer-facing or finance workflows. Keep a "human in the loop" for critical tasks like client correspondence, billing and regulatory reporting.
• Treat AI policies as standard operating procedure: Incorporate generative tools into the same quality control and risk management frameworks you use for normal staff workflows. Clear guidelines give your team the confidence to embrace automation safely without hiding errors.