Australian APRA-regulated super funds manage more than $3 trillion in retirement savings through governance frameworks designed before dynamic asset allocation, private credit co-investment, and real-time geopolitical repricing became standard features of the investment landscape.

Those frameworks aren’t broken, but they weren’t built for the speed at which material investment decisions now need to be made. Artificial intelligence now offers a practical and lawful way to close that gap, without requiring legislative change, new APRA guidance or a renegotiation of the Superannuation Industry (Supervision) Act (SIS Act).

What the law actually requires

The starting point is the prudent person standard under section 52 of the SIS Act. The standard is outcome-oriented, not process-prescriptive: it does not specify how a trustee should gather information, only that the decision reflects prudent care and diligence. It is also not frozen in time. In fact, a fund that consistently ignores analytical tools its peers are adopting will find it increasingly difficult to argue its process meets the standard.

That reading is reinforced by APRA’s SPS 530, which requires RSE licensees to formulate and maintain an investment strategy addressing risk, return, liquidity, and diversification, subject to ongoing review. A governance framework that structurally inhibits timely review – because information takes days to prepare and committees meet monthly – sits uncomfortably with that obligation, particularly in volatile market conditions.

None of this creates a legal obligation to deploy AI. What it does is confirm there is no barrier to doing so, and that the direction of regulatory expectation is toward more rigorous information processing, not less.

The delegation framework is already there

One answer lies in the delegated authority framework under section 59 of the SIS Act, which is among the most under-utilised tools in super fund governance. Trustees may delegate defined investment authorities to the CIO or investment committee, but many delegation matrices were designed for a world where the primary question was which asset manager to appoint, not whether to act on a rates shock before the close of business or accept a co-investment with a 72-hour decision window.

The fix does not require touching the law. It requires reviewing the delegation matrix so that time-sensitive decisions within defined risk parameters have a clear authorised decision-maker, supported by fit-for-purpose information infrastructure. An AI system that synthesises the fund’s positions, risk budget headroom and relevant market context in real time is precisely what makes delegated authority function at the speed modern markets demand.

CPS 230 and the operational resilience argument

APRA’s CPS 230, which took effect in July 2025, introduces an explicit operational resilience framework for APRA-regulated entities. The standard requires funds to identify their critical operations, assess associated risks and maintain the capability to deliver them through periods of disruption.

Investment decision-making is, without question, a critical operation for a superannuation fund. CPS 230 requires that the people, processes, technology and data supporting critical operations be fit for purpose and resilient. This creates an affirmative obligation to consider whether the current information infrastructure supporting investment decisions is adequate – not merely for normal conditions, but for the speed and complexity of environments in which material investment decisions now need to be made.

APRA has signalled clearly that it expects regulated entities to leverage technology and data capabilities to enhance decision-making. Combined with the business continuity obligations under SPS 232, the CPS 230 framework gives the risk and compliance function a strong platform to champion AI investment, rather than treating it with reflexive scepticism.

What AI can lawfully do in the investment process

With the legal framework established, the practical question is what AI can actually do inside the investment governance process. Four applications stand out as clearly lawful, immediately implementable and genuinely value-additive.

The first is pre-meeting synthesis. An AI system grounded in the fund’s knowledge base – positions, SAA policy, risk budget, prior decisions – can produce a structured briefing oriented to current market conditions before each committee meeting, which would be available in real time, consistent in quality and not dependent on any one person’s bandwidth. The trustee still reviews information and makes the decision; the information is simply better and more timely.

The second is decision consistency checking. An AI system can flag in real time when a proposal diverges from the fund’s stated investment beliefs, SAA policy or risk appetite – surfacing relevant context so the committee can make a fully informed choice, without replacing trustee judgement.

The third is real-time scenario analysis. When a material market event occurs – say, a rates decision, geopolitical shock or significant equity repricing – an AI system can model portfolio implications in minutes rather than hours, making it possible to act within a delegated authority framework on the day the event occurs rather than at the next scheduled meeting.

The fourth is live decision support during meetings. Some of the world’s most sophisticated long-term asset owners, including sovereign wealth vehicles in the Gulf, already deploy AI advisers that surface relevant data and prior decisions in real time. Under Australian trustee law there is no impediment to a similar model. The AI is an input. The trustee remains the decision-maker. The fiduciary duty is entirely unaffected.

The case for moving now

Governing AI systems well is not a new kind of challenge, it is an existing challenge applied to a new category of service. The conflicts of interest obligations under SPS 521, the outsourcing framework under SPS 231 and CPS 234’s information security requirements provide a ready-made template: assess the vendor, classify the data, maintain human oversight and document the process. These are the same disciplines applied when appointing an investment consultant or custody bank. Getting the governance of AI right is not a reason to delay, but a reason to start.

Australian superannuation has produced world-class investment outcomes over three decades. Its architecture – the trustee model, the profit-to-member mandate, the regulatory framework – deserves credit for that. The question is not whether the architecture is sound. It is whether the operational infrastructure inside it is keeping pace with the environment in which it now operates.

The APRA performance test sharpens that question with each passing year. AI offers the investment committee something it has always needed but rarely had in full: the ability to be comprehensively informed at the moment a decision needs to be made, rather than at the moment the calendar happens to allow. That is not a challenge to trustee governance, but a natural evolution of it.

The law supports it. The regulatory framework reinforces it. The competitive environment increasingly demands it. The funds that move first will not merely make better decisions in the short term, they will build the institutional capability that defines best practice for the decade ahead.