UniSuper starts to explore agentic AI – Cloud – Software

UniSuper is starting to explore opportunities with agentic AI, with an initial use case aimed at helping a financial advisory team take notes and serve more clients.

Speaking to the iTnews Podcast, senior manager of emerging technology and efficiency Matt Bright said the fund is using two Microsoft tools – Copilot Studio and Power Automate – to underpin its agentic AI work.

The first agent assists a team of 50 comprehensive financial advisors to summarise conversations with fund members.

Where they once took notes by hand, the notes are now automatically captured.

“At the end of the meeting, [the agent] takes the transcript, populates a file note template, and then the advisor does quality assurance to make sure it’s accurate,” Bright said.

“That might take them five minutes, saving them 25 minutes for every member interaction over the course of a year.”

UniSuper estimates about 2000 hours of time is saved per year with this single automation, and that meant the team could do about seven percent more meetings in the same time.

“Our advice file notes solution is built in Copilot Studio, but it weaves Power Automate into the workflow,” Bright said.

“I think that’s what we’re seeing more and more is that those two [tools] work hand-in-hand.”

Bright said further use cases are being designed by working directly with Microsoft as well as some “strategic partners”.

“We’ve got quite a few initiatives in place to extend and further double down on our Copilot investment.”

Copilot to all staff

UniSuper joined the early access program run by Microsoft to access Copilot, starting with an initial user base of 30 over four weeks, before expanding the trial cohort up to 300 – about one in four staff.

Trialists were drawn from across internal functions, but tended to be heavy M365 users

“We wanted people that send a lot of emails that do a lot of reporting that are in Teams that are having calls,” Bright said.

To build internal confidence, UniSuper ran a formal training and adoption program with Microsoft certified partner support, and offered “constant, repeat” training to staff; set up a community in Teams to share knowledge, prompts and answer questions; ran drop-in sessions; and offered more confidential channels “to log issues or concerns”.

The repeat training was necessary in part because Copilot continued to evolve even during the trial phase.

To aid with the pace of change of the technology, UniSuper offered up a “60-to-90 second” video every Friday to communicate new or perhaps under-utilised features or capabilities of Copilot.

“People had a constant reminder of ‘Have you tried this capability? Have you thought about this?’” Bright said.

Prompt writing was also a particular focus area for upskilling.

“The key to using foundational tools like Copilot is that if you ask good questions, you get good outcomes as long as the information that it’s running on is solid; but if you ask simple questions, you generally get simple responses,” Bright said.

“We put a lot of effort into prompt writing and [provided] prescriptive guidance on how to write and structure good prompts as well.”

UniSuper expanded Copilot access to all 1200 of its staff in September last year.

“In 2025, we’re hovering at about 80 percent active utilisation, which for an organization of 1200 people is a really strong rate and something we’re really encouraged by,” Bright said.

The fund had also seen a corresponding decline in the use of public-facing AI platforms and tools.

“They’ve gone down quite significantly because our people are using Copilot,” he said.

Member experience

While use cases so far are back office focused, UniSuper ultimately sees AI, and generative AI, as having a role to play in improving member outcomes.

“UniSuper has really clear and deep ambitions on wanting to be a leader in retirement,” Bright said.

“Our CEO Peter Chun often speaks about the fact that we look at ‘the Netflix of retirement’ as our motto, and that essentially speaks to the fact that we want people to have personalised experiences that are tailored to their needs.

“If you look at the retirement lifecycle, we want to make sure that we’re interacting and engaging with you in a way in which it sets you up for the best possible outcome in retirement.

“We think generative AI is going to have a really big role to play in how we both understand and engage with our members and equally how we increasingly allow and enable digital self-service through our member online and mobile app platforms.”