The 2024 Australian Responsible AI Index has revealed a significant gap between the perception and implementation of responsible AI practices in Australian businesses.

The report by market research agency Fifth Quadrant and sponsored by the National Artificial Intelligence Centre examined the way Australian businesses are adopting responsible AI practices to gain insight into the key areas of progress and improvement.

The research found that while 78 per cent of businesses agreed with the responsible AI statements provided to them, only 29 per cent had implemented these practices into their business operations.

Fifth Quadrant director Steve Nuttall said businesses may struggle to take high-level ethical AI principles and put them into practice, resulting in this 49 per cent perception gap.

“I don’t think they have been exposed properly or fully to the detail that’s involved in building out and executing AI in terms of the practical steps that they actually need to go through,” he said.

“There are resources there but now it’s just making organisations aware that it’s there.”

Dr Nuttall said it was important for organisations to adopt the Voluntary AI Safety Standards and engage with the guardrails in order to narrow this gap.

Respondents to the survey were scored on their implementation of 38 responsible AI practices, with more practices implemented resulting in a higher index score. The mean Responsible AI Index score was 44 out of 100, with organisations on average only implementing 12 responsible AI practices.

A common characteristic found among organisations with higher responsible AI index scores was business leadership playing an active role in driving AI development.

Dr Nuttall said: “Ultimately, it all starts with leadership and having the right governance in place and building that from the ground up.

“The businesses that are developing AI applications that don’t have much support from the leadership, they don’t have the right governance in place, they’re more likely to be the ones that may cause harm and I think that represents a level of risk to the business.”

Irresponsible use of AI can present legal and reputational risks for businesses. AI systems have the potential to create unwanted bias if not implemented correctly, which may breach anti-discrimination laws.

Businesses also risk reputational damage and losing consumer trust if they are not open and transparent with their AI usage.

The Future Workplace: Opportunities & Adversities in AI report commissioned by Persol Kelly found there is some hesitancy surrounding the implementation of AI.

The research found that 40 per cent of Australian businesses were reluctant to adopt AI practices. Several barriers to AI adoption were identified, including lack of resources, lack of knowledge and concerns over risking customer trust and brand reputation.

Dr Nuttall said organisations that have not adopted AI or are not aware of the potential of AI have completely underestimated the value AI can bring to their business.

“I think a lot of companies feel that AI is moving too fast, and they don’t know when to invest,” he said.

“If you’re doing AI responsibly, ultimately once you’ve got those frameworks and practices embedded into your organisation, there are short-term costs but real long‑term benefits. You build out more trust with your customers, you build better products, you build out AI solutions that are more likely to succeed than fail.

“It’s a complete misconception that responsible AI is not giving you a competitive advantage, it does.”

Despite the hesitancy that exists, many businesses who have adopted AI practices into their business operations have found it to be of benefit.

Digital marketing agency White Chalk Road’s strategy director Alex Hamilton said AI has provided benefits and played a role in both internal and external business operations.

Mr Hamilton said AI has saved time by carrying out mundane tasks that are easy to do with AI and assisted in creating consistent communication.

Mr Hamilton highlighted the importance of AI being used to assist the user rather than replacing them altogether.