Westpac is scaling its AI Accelerator, the banking group’s dedicated artificial intelligence (AI) innovation incubator and high-speed development hub, to deliver a continuous stream of ready-to-deploy AI and generative AI (GenAI) technologies for real-world use cases.

Presenting at the Future of Financial Services 2024 conference on Friday, Westpac’s chief technology officer (CTO) David Walker underscored the bank’s ambition to build a ‘true copilot’ system – imagined as a full-service, personalised digital consultant or concierge for staff and, eventually, for servicing customers.

“Copilots are going to change the way we think about how banks interact with people. They’re always on, they’re highly personalised, and [have a] higher level of intelligence.

“Copilots will be able to speak to you, support you [and] give you clear advice, and go through and do things for you on your behalf.”

According to Walker, Westpac’s AI Accelerator, quietly launched by the bank earlier this year, has embraced a ‘platform-based’ design approach, leveraging a set piece of reusable “patterns” to fast-track development of AI tools that address specific areas of need, whilst strictly following its responsible AI framework.

In Westpac’s case, these ‘patterns’ were developed through an exhaustive staff consultations process, with staff providing feedback on process pain points as well as delays or roadblocks they face in their day to day tasks.

“We went out and spoke to a large group of our communities [staff] who told us of all these different use cases. We collected them and built stories about what they’d look like.”

The AI Accelerator team, working with this feedback, categorised these demands into four defined categories of need:

• Information retrieval – For instance, using natural language search to unpack answers from voluminous datasets or documents.
• Content generation – The ability to create any type of new content (e.g., communications, reports, code etc.) in real time.
• Process completion – Simplifying processes and workflows; “removing the toil from everyday tasks”.
• Data insights – Empowering staff to analyse data and unlock insights​.

“These are really powerful ideas, because if you can get these patterns right, then you can get them to be repeatable,” Walker said.

“We can build a platform to support these use cases, then anyone could turn up and with these problems that they might have or these opportunities they might have then they’re able to scale them.”

Among the real-world AI and GenAI use cases developed through the AI Accelerator (based on each of the four ‘staff demand’ categories), include:

• a ‘mortgage specialist’ virtual assistant – a next-gen chatbot that promises, Walker said, “to do what early generations of chatbot couldn’t”.Trained “in just a few hours”, he said, ingesting significant volumes of documents that describe the mortgage process and Westpac’s credit rules. The chatbot, used internally, can be asked and respond to “virtually any question” regarding the Westpac mortgage process.
• an AI pair programmer – already live within the bank’s engineering team, Walker reports that the tool has delivered a 10-25 per cent increase in software code generated each day by Westpac developers.
• a mortgage assessor AI – focused on brokers, this “powerful” tool automates and fast-tracks the loan documents assessment process, including “handwritten notes and filled-in forms”, to determine applicants’ eligibility for home loans.“We triage every one of the documents [submitted to us by brokers], which is typically a very human step in the process.”What we’re able to do [with this AI] is take that data in – which often includes handwritten notes and filled-in forms – assess those, and put a recommendation back to a human inside a bank. We then pass that back to a broker, which speeds up the [application] process dramatically.”
• risk and compliance AI – developed to manage risk across the bank, this tool looks for patterns in all internal data held around risk.

While useful in their own right, for Walker, these tools serve as a stepping stone to a long-term ambition of building comprehensive copilot systems – ones that “truly enables the encapsulation of knowledge”.

“It’s the expertise layer that will be the real game changer here.”

He concluded: “We believe Westpac is all about relationships. And we believe that AI deepens relationships, it doesn’t replace them. We see that in everything we do.”