AustralianSuper bets on generative AI, workforce upskilling

AustralianSuper is pressing ahead with generative AI deployments for investment and back-office staff, undeterred by lacklustre productivity gains from initial trials.

AustralianSuper is pressing ahead with generative AI deployments for investment and back-office staff, undeterred by lacklustre productivity gains from initial trials, chief technology officer Mike Backeberg says.

The $315bn super fund is positioning itself as a leader in AI adoption and believes fostering an AI-adept workforce is paramount to extracting maximum efficiencies and amplify value for its over 3.3 million members.

“The vision that we have is that artificial intelligence is a technology which is going to influence and change the world, and the opportunities right now are tremendous,” Mr Backeberg tells The Australian.

Many financial services workers are demanding access to the technology which is tipped to contribute up to $115bn a year to Australia’s economy by 2030. However, a disconnect is also emerging between those seeking to turbocharge their jobs and those who distrust the tech.

The country's largest pension fund is rolling out a comprehensive education program to help its over 1500-strong workforce get up to speed through workshops, on-the-job training, and AI “champions” in each department.

Almost six months ago, AustralianSuper was one of a number of early adopting companies piloting generative AI tools built by Microsoft. The program had a “modest” productivity target rate of 5 per cent, which Mr Backeberg said was not achieved.

“We wanted to see a 5 per cent increase in efficiency or productivity across the organisation. And that equates to about two hours per week per colleague,” he told The Australian.

“What we’ve been able to measure through the adoption program is about an hour a week gain in productivity across the 260 colleagues who have used that program.”

Some individuals and teams had much higher gains. For example, software development in its information security department had targeted a productivity boost of 25 per cent and had achieved north of 20 per cent.

Microsoft’s Copilot is now available to all staff and training should further boost productivity as workers familiarise themselves with how they can use it on day-to-day tasks.

AustralianSuper was also piloting two key “primary use” case studies. The first involves using generative AI tools to “digest”, analyse and “curate” data, even if unstructured, from investments and research.

“When you are looking at investments in assets or managing an existing asset, there’s a lot of information that is filtered through those teams. AI is an opportunity for us to digest that, to analyse it, and to provide insights to those research teams and investment colleagues,” Mr Backeberg said.

He warned, however, the experience so far showed the technology still needed further finetuning before it could be rolled out across its investment teams in the US, the UK and Australia. This is mainly because error rates were still unacceptably high.

“I want to say that we’ll do it in the next 12 months, but what we need to be able to demonstrate is we’re getting the value out of it. Because I have seen examples of where people have been using AI to do analysis on unstructured data, and it’s been generating results which are incorrect,” he said.

Given the fund handles the retirement savings of over 3.3m people, “we cannot have that (errors) emerging in our environment,” Mr Backeberg said. “We’re not talking about small amounts of money.”

Mr Backeberg said AustralianSuper didn’t have a specific budget allocated for AI projects, but the technology would likely be embedded in most of its future technology investments.

A second primary use case the fund is exploring includes using generative AI to improve member services and experiences, such as through first contact resolution.

It plans to offer such tools to back-office staff first, and then assess, as the technology and regulations evolve, whether it can be provided to members directly.

“We need to temper the speed at which we do (this) — absolutely we can get value internally quickly — but from a member perspective, we want to be absolutely certain members are safe, secure and that they can actually trust the service that we provide,” he said.

Acknowledging the fund hasn’t surveyed employees to gauge whether its 1500-plus staff viewed generative AI as a job threat or a productivity booster, Mr Backeberg said management sees it firmly as the latter.

“We do not want people to have a sense of fear of AI,” he said.

“We are very clear that we do not see AI in its current form replacing people’s roles, not in AustralianSuper. We see AI as really enabling colleagues to focus a lot more on high value tasks.”

The fund is looking to add 25 roles in its IT department in the next 6 months, growing it 12 per cent to 229 people.

“We continue to grow as an organisation, and there’s a scarcity of skilled people in industry across every area that we work in, so for us AI at the moment is most certainly not about replacing people’s roles.”

Almost two-thirds of Australians distrust artificial intelligence, according to human resources software titan Workday. \

Another survey by Veritas Technologies found a large proportion of local workers had concerns around security, accuracy, and compliance issues when using generative AI.