CBA turns focus on applying responsible AI following national ethics framework pilot

Commonwealth Bank of Australia plans to leverage AI at scale, but it wants to ensure that it's being done in a responsible and ethical way.

Commonwealth Bank of Australia (CBA) has revealed that testing the federal government's artificial intelligence (AI) ethics principles during the creation and design of its Commbank app feature Bill Sense gave insight into how the bank could apply responsible AI at scale.

"It was great to see that the AI principles sat really neatly with the control and governance frameworks that the bank already had in place. Things like the safe management of data, customer privacy, and transparency have been central to the way we operate since long before the advent of AI," CBA chief decision scientist Dan Jermyn told ZDNet.

"But the pace, scale, and sophistication of AI solutions mean we need to ensure we are constantly evolving to meet the demands of new technology, which is why collaboration with our partners across government and industry is so important."

The bank was one of six organisations that volunteered to trial a series of eight AI principles that were developed as part of the national AI ethics framework in real-world scenarios.

The eight ethics principles that were developed for the framework included: Human, social and environmental wellbeing; human-centre values in respect to human rights, diversity, and the autonomy of individuals; fairness; privacy protection and security of data; reliability and safety in accordance to the intended purpose of the AI systems; transparency and explainability; contestability; and accountability.

The ethics principles were developed following the release of a discussion paper earlier this year by Data61, the digital innovation arm of the Commonwealth Scientific and Industrial Research Organisation (CSIRO).

Jermyn explained running the pilot when the bank was developing Bill Sense was a logical decision.

"Bill Sense is powered by 20 machine learning models, processing 16 billion transactions, and so it was a great case study for how we think about responsible AI at scale," he said.

"We looked at it through the prism of each of the eight AI ethics principles, ensuring that we were adhering to both the letter and spirit of the guidelines. This was a complete end-to-end process, from the ideation of the features, right through to customer experience and feedback."

But it is not only Bill Sense where the bank is leveraging AI technology, the use of AI across the bank is broad, Jermyn said.

"AI is a really important tool for us in making sure we can provide personalised experiences that improve our customers' financial wellbeing, and is widely used to improve our processes," he said. "To give some idea of the scale, our customer engagement engine uses 400 machine learning models, that make decisions across 157 billion data points."

In a bid to further ensure AI is being applied responsibly, the bank has developed a tool to make it easier for teams across the bank to deliver AI safely to scale, according to Jermyn.

"For example, we have developed 'explainable AI' capability, which makes it simple for any of our business teams to understand and explain the key drivers of even the most complex deep learning models," he said.

He added that responsible application of AI will be necessary as CBA continues to use it as a tool to improve customer experience.

"We see AI as a key enabler for us in providing a great, personalised experience to all of our customers, and so we are committed to ensuring we apply it in a consistently fair and ethical manner," Jermyn said. "As we continue to grow the ways in which AI helps us to support the financial wellbeing of our customers and communities, it's essential that we do so in a responsible and sustainable way."

Related Coverage