AI won’t ever make our investment decisions: State Super CEO

One of Australia’s oldest superannuation funds is integrating artificial intelligence into its investments process, but the fund’s people will always be firmly in the driver’s seat.
AI won’t ever make our investment decisions: State Super CEO

State Super is embracing the artificial intelligence revolution by integrating machine learning aimed at enhancing its investment team’s decision-making, not to replace it, according to John Livanas, the fund’s chief executive.

John Livanas,
State Super

“We didn't set out with the goal of implementing AI or machine learning, but we approached it by looking at the problem we needed to solve,” Livanas told AsianInvestor.

State Super manages a complex A$38 billion ($25 billion) portfolio and funds-management process with a small team. As a closed defined-benefit fund, it must also contend with negative cash flow.

“These challenges mean we need to be proactive and minimise potential downsides by being highly responsive to market movements,” he said.

ALSO READ: State Super picks partner to manage alt risk premia portfolio

Along with Chief Investment Officer Charlie Wu, Livanas saw that the volume of data the investment team encountered on a daily basis was overwhelming.

“Up to 2,900 data points, which is beyond human capacity to process effectively,” he said. “This led us to initiate a data science project to make sense of this data.

About two-and-a-half years ago, the State Super team began to construct reinforcement learning models.

“These models analyse the data, discern correlations, and ascertain predictive values, both directly and relatively. This marked the beginning of our AI journey,” Livanas said.


In their attempt to integrate AI responsibly, Livanas and Wu implemented a robust governance framework to oversee adoption of the technology along two key components.

First, they formed a governance panel with experts from machine learning and data science backgrounds.

“This panel's role was to oversee model selection and implementation strategies, ensuring practices like version control, data residency entirely within Australia, and comprehensive documentation policies were in place,” said Livanas.

Secondly, the fund developed policy statements to emphasise that AI tools would support – not replace – investment decision-making.

“Human oversight remained integral, ensuring a balanced perspective for our investment team to evaluate market data effectively,” he said.

Also read: 2024 predictions: AI will reshape asset management hiring needs

“This approach provided me with a sense of security because it reinforced the idea of using tools to enhance our investment team's capabilities without introducing biases,” the chief executive said.

State Super’s investment professionals now utilise reinforcement-learning and self-learning models, which fall under the broad AI umbrella but are more specialised.

“Considering the value these tools have added, it's evident that combining human expertise with sophisticated models yields better outcomes than relying on human analysis alone,” said Livanas.  

State Super has since observed outperformance in its investments that was attributable to these tools, including where the value gained significantly outweighed costs. These gains justified the investment from a cost-benefit standpoint.

“The robustness in data management and model construction has enhanced our decision-making process,” Livanas said.


Elsewhere in the investment world, JP Morgan Asset Management has invested significant time and effort into harnessing the power of AI for its global investment processes, focusing on generative AI, according to Lee Bray the firm’s head of equity trading for Asia Pacific.

Lee Bray,

Like State Super, the global asset manager is putting its faith in human intelligence to make the final decisions.

“The key for us in the AI space is not to replace human judgement, the skills of our investors are the ‘secret sauce’ that make JPMAM so successful,” Bray told AsianInvestor.

ALSO READ: Institutions embrace AI for portfolio optimisation potential

“Our aim is to make the investment process more efficient, so we’ve focused very heavily on a collaborative approach where the ideation process for AI products sits with the investors rather than away from them,” he said. “This ensures we develop AI related products with a focus on adding value to investors rather than replicating them.”

Asset managers at the firm have been developing and testing a generative AI tool called SpectrumGPT alongside internal investment professionals.

“With this cutting-edge AI tool, our analysists and portfolio managers will be able to work with and make sense of massive amounts of unstructured data, from internal research reports to company call transcripts, to help them make informed investment decisions faster,” he said.

Throughout 2023 and into 2024, JPMAM’s focus in terms of AI adoption has been on the equity investment process. There are plans into the future for similar products in the fixed income and multi asset world, according to Bray.


¬ Haymarket Media Limited. All rights reserved.