Suncorp Group is planning a multi-agent approach to its AI strategy, building a library of reusable components to help scale the technology across its brands.
The insurer, whose brands include AAMI, Gio Insurance and API, said in October it had begun plans to transform its customer service and claims process.
More recently, at Databricks' 2025 Sydney event, the group's machine learning chief engineer Tooraj Valai said the business case for expanding the use of AI is strong, having already leveraged AI to save thousands of man hours and generate more than a million words in case summaries.
“All of this points to us being very good at building AI companies… But to be honest, here’s where we see our future. [that it’s] It's not just about having one high-value, discrete AI application.
“It's mainly about how these applications and AI and agents can work together and interact with each other, and how we can build complex multi-agent systems,” Varaee said.
The group has therefore adopted a system architecture comprised of the Databricks suite of products, which Varaee says is tailored to a multi-agent approach.
The suite consists of Databricks Lakehouse for data management and Unity Catalog for governance.
This architecture provides the foundation for Databricks' enterprise suite for building, delivering, and releasing generative AI applications called Mosaic AI.
Mr. Varaih said that given the regulatory and governance hurdles that prudential authorities place on insurers, as well as the technology that the insurance industry uses, it is not practical to reinvent a new agency from scratch for each business case.
“The next layer is a scalable engine for us. You can imagine that if you want to incorporate agents into an enterprise company, you can't start everything from scratch and build everything from scratch.
“This is not scalable. This is not regulated. Our approach [to] This work builds a layer consisting of all reusable components and services so that all agents [will] you will be able to use them [in a] A unified and integrated method.
“It consists of an observability layer, reusable agents, components, and agent context memory that provides stateful functionality for agents.
“And all these accumulations [will] You will be able to empower agents that you build in the future. “This is the beauty of this architecture,” he said.
“We're going to have plug-and-play functionality here. We're going to be able to use previously built agents.”
Varaee provided an update on Suncorp's achievements in leveraging AI since the company announced its program of initiatives in October.
He said the insurance company's claims department was using AI to automatically generate case summaries of more than 1 million words.
He also revealed that the group's Search Augmented Generative Base (RAG) Generative AI Smart Knowledge application has saved customer service staff more than 15,000 hours of work time.
“For us, AI is not just an interesting R&D project; it is the primary engine for scaling responses to handle claims. [an] exact method [while] We also do risk management,” Varaih said.
