Credit Union Australia set out on the AI path to make life easier for customers and employees. Here are CUA's lessons to help boards make the most of AI.
CUA is Australia’s largest credit union, offering finance and insurance services to 500,000 customers. Based in Brisbane, it has 950 staff, call centres in Sydney and Melbourne — and an AI-infused chatbot called Sam.
After an initial successful trial of Sam, (based on Australian-developed Flamingo AI technology) who helps CUA Health customers through their health insurance processes, the CUA board has extended the pilot to June.
According to CUA head of digital innovation Melissa Witheriff, it’s more efficient, typically cutting application times from 30 to seven minutes. Sam collects real-time insights for the business using staff know-how.
“Our board are incredibly interested in the application of new technology,” says Witheriff, adding directors need to “ask the right questions — is there real benefit or is this just interesting technology?”
CUA chair Nigel Ampherlaw says the board has taken time to explore the issues and is developing an AI/robotics strategy.
In 2018, the CUA board decided that to survive, given its size, it had to do something different. After a board trip to the US, which included time in Silicon Valley learning about AI, “we concluded we had to spend 18 months to become a better bank and beyond that, better than a bank”.
“We came up with the ethos that you put people before profits,” says Ampherlaw. “When it came to AI robotics, we had two guiding principles. On the customer side, any robotics/digital we did was to enhance the quality of human interaction, not replace it. From the employee side, we weren’t going to use it to cut costs and heads.”
One example is the frustration of customers dealing with call centres. “People call in, they never speak to the same person twice and a lot of the subtleties are lost in translation,” he says. “Call centres are a source of much customer dissatisfaction.”
CUA now uses an instant messaging app integrated with its online banking that allows customers to nominate their own personal banker. From then on, every interaction with CUA is with this one person.
“The current phase is using robotics to automate tasks,” says Ampherlaw. “AI is a way down the road. On the horizon is automation and teaching robots to do certain tasks.” v
The board asked the head of HR to research the issues, working with a US consultant. Board and management needed to understand how automation would affect key jobs and the skills required. HR identified the digital capabilities and potential of its workforce. Most fell into two categories: digital natives (who had grown up in a digital world) and digital migrants (who had adapted). The third category were digital refugees who CUA wanted to find opportunities to redeploy.
“This is not a headcount-reduction exercise, says Ampherlaw. “We want to use robotics to replace mundane, repetitive tasks; to redeploy those people into customer-facing tasks. We have a robot that checks IDs, and risk and compliance have also automated activities.” He says it is important to build a solid risk management framework. For example, the person who designs something is not the person who checks it. And robots making decisions is still a way off. “From a customer point of view, we’ve started from the importance of human interaction to build trust.”
The board asked chief operating officer Brigid Leishman GAICD to develop a robotics strategy with strong governance processes around it and CUA is planning a half-day session for the board in understanding bias.
“One of the big concerns is unconscious bias when you’re teaching a machine to make decisions and judgements,” says Ampherlaw. “If you teach a robot to do the wrong thing, it becomes a systemic risk. From a governance perspective, we take a deep interest in this. It was the board that led the charge on AI. The board is naturally curious because of its diversity.”
Questions for boards
- What is our company’s AI strategy?
- What AI or shadow AI (AI that is being done without the knowledge of the CIO or CTO) is there already?
- Who will be the head of AI and who will coordinate all types of AI being introduced?
- Does AI sit with technology, the CIO, digital, transformation, innovation or another department?
- How will we ensure the algorithms that are being used by our organisation are accurate and the right models for our business?
- Should we build AI in-house, or engage one vendor, or select the “best of breed”?
- Who has trained or is training the algorithms?
- How will we ensure that the algorithms have not been trained with bias?
- What are our security requirements around AI?
- Given AI will replace or augment human jobs, what will be our future organisational structure that will include digital labour positions as well as HAVA (human assisted virtual assistants) or HAMA (human-assisted machine-assisted) roles?
- How do we develop a culture that embraces human-machine workplaces in order for the technology to amplify legacy system capability and augment human capability?
- How will we manage communications to employees about AI and how it might affect their employment?
Source: Flamingo AI