The Commonwealth Bank says using Artificial Intelligence to help protect its customers has reduced fraud by 30% and seen a 50% reduction in scam losses.
Speaking at the AWS Data and AI Day in Melbourne on March 4, the bank’s Chief Data and Analytics Officer, Dr Andrew McMullan, said the use of AI models is not only helping to predict fraud but also enabling the bank to alert customers quicker and more frequently than ever before.
“We have had a real focus over the past few years to do everything we can to protect our customers including moving some of the traditional fraud rules on to more advanced AI prediction to protect our customers and communities in partnership with AWS,” he said.
Dr McMullan said what had fast-tracked CBA’s development was the launch in September last year of its cutting-edge AI Factory. This leverages Amazon SageMaker, a machine learning managed service, and Amazon EC2 P5 Instances for deep learning and high-performance computing applications. The P5s leverages NVIDIA chips to perform AI-powered tasks four times faster and 40 per cent more efficient than the previous set up.
“This has really enabled us to accelerate and continue to deliver on our ambitious AI goals,” he told the audience of nearly 1,000.
“The AI Factory is available to everyone in the enterprise. It has empowered all of our engineers and scientists to embed AI and generative AI into their day-to-day activities. This in turn has seen increased creativity and output across the organisation by providing all of our people the ability to use generative AI in their jobs.”
Dr McMullan said the AI Factory has accelerated the bank’s development and AI powered initiatives by around four times compared to the previous rate.
“It has also enabled us to create what we call our generative AI studio where we allow all of our engineers and AI practitioners to have access to reuseable assets and protocols. This includes Anthropic’s latest models – Claude 3.5 currently and soon to be 3.7 – so that we can use the best models available globally to make us much more creative and productive.”
Dr McMullan also spoke about what CommBank was doing to make sure it safely scaled and used AI and generative AI responsibly.
“In early 2023 when open source and other large language models became available, we realised that the combined power of generative AI and predictive AI was going to create new opportunities and new ways of working across the whole enterprise.
“So, we undertook an enterprise AI acceleration program. One of those work streams was talent and skilling.
“Distinguished engineers have been a very successful resource at CommBank for several years and off the back of that we wanted to create the equivalent role for a Distinguished AI Researcher or a Distinguished AI Scientist. We have hired seven Distinguished AI Scientists and are also promoting our best people and deploying them in every part of the organisation.”
Dr McMullan said the bank also looked at its Risk and Governance work stream. “We did a review of all of the processes across the organisation to ensure we had the right frameworks and the right way of governing and observing everything that we were doing with AI and generative AI across the group.
“We had an accelerated data foundation program where we knew that the data is the fuel for AI and generative AI. This meant accelerating our AI migration onto AWS. The initial plan was that it would take us 18 months to migrate all of our data platforms, but we will deliver it in nine months, and we will go live in May.”
Last month CommBank announced it had entered a five-year strategic collaboration with AWS to continue as the bank’s preferred cloud provider.