It’s an exciting time. You’ve identified an AI use case, you’ve identified the benefits, and now you’re building the business case.
As you do so, you’re laser-focused on one question: “What would give the board enough confidence to give the go-ahead?”
If you’re an APRA-regulated entity, its April 2026 Letter to Industry on Artificial Intelligence is a good place to look first. It covers both the opportunity and areas of concern associated with AI rollouts:
- Opportunity: APRA calls out that failing to embrace AI may put businesses at a strategic disadvantage.
- Area of concern: Boards need sufficient understanding to challenge whether risks such as unpredictable model behaviour and the impact on critical operations have been properly examined.
You’ve probably got the first one covered. To address the second, you should have done the groundwork needed to give the board evidence it can use to meet APRA’s expectations.
An example
Imagine an insurer proposes to use AI to read incoming claims, identify missing information and prepare a summary for an assessor.
You can come up with quite a few benefits, but let’s look at how you could address the concerns by asking questions like these:
- Can the system contact the customer?
- Can it update the claim record?
- Can it recommend an outcome?
- Which actions require someone to check first?
Depending on the answers, you’ll need to bring evidence that addresses questions such as:
- What is the system allowed to do, and where must a person intervene?
- How has it performed on representative cases, including those with incomplete or contradictory information?
- What happens when it makes a mistake or becomes unavailable?
- Who monitors it after launch, and what would trigger a pause or review?
A note on APRA’s expectations
APRA’s expectations cover several areas, including accountability across the AI lifecycle, proportionate ongoing monitoring and credible fallback processes where AI supports critical operations. It also calls for visibility into suppliers and their underlying third- and fourth-party dependencies.
Source: APRA Letter to Industry on Artificial Intelligence (AI)