In this ADAPT panel, Matt Boon is joined by Elizabeth Tidd, Australian Information Commissioner and Charles McHardy, CIO and Digital Officer at Services Australia. They examine how government can unlock AI productivity while safeguarding transparency, privacy and trust. The discussion positions AI as both a performance lever and a governance challenge.

Proving value: the AI dividend must be visible

Matt argues that organisations must earn the AI dividend through clear, measurable outcomes. AI cannot remain experimental. Leaders must demonstrate improved efficiency, faster service delivery and tangible impact.

Charles supports this with practical examples. Services Australia uses AI to speed up claims processing, improve call handling and increase delivery capacity within existing budgets. However, Matt highlights a barrier. Many stakeholders still struggle to understand AI, which slows adoption and weakens executive confidence.

Trust defines adoption

Elizabeth positions trust as the central constraint. AI lacks human cues, which creates inherent scepticism. Transparency and accountability therefore become essential. She highlights strong contrasts in public sentiment. Trust in AI companies remains very low, while trust in government is significantly higher. This creates an opportunity for the public sector to lead. However, this trust is conditional. Citizens expect clear notification when AI is used, transparency in data sharing and strong privacy protections.

Charles reinforces this through MyGov. Citizen trust determines whether people engage with government platforms. Without confidence in security and clarity, adoption declines.

Governance enables progress

Elizabeth challenges the idea that governance slows innovation. Governance provides control, builds trust and enables safe scaling. Without it, organisations risk failure and public backlash.

Charles operationalises this by implementing structured AI assurance frameworks. These assess risk, define acceptable use and align legal, procurement and cyber functions. Governance becomes a practical tool that supports delivery rather than blocking it.

Drawing clear risk boundaries

The panel distinguishes between supporting decisions and making them.

Charles sets a clear boundary. High-risk use cases such as fully automated decision making sit above the acceptable threshold. Services Australia prioritises AI that supports staff, such as preparing documents or improving knowledge access.

Elizabeth explains why this matters. Many agencies fail to disclose their use of automated decision making, which undermines trust. Organisations must clearly communicate where AI influences outcomes and ensure meaningful human oversight.

Focus on practical use cases

Charles identifies three areas that drive productivity:

• knowledge systems that improve staff decision support
• optimisation of resources and workflows
• faster software development.

These use cases enhance existing processes rather than replace them. This allows organisations to deliver value quickly while managing risk. Elizabeth notes that lower-risk applications, such as accessing standard procedures, can improve accuracy and consistency without compromising privacy.

Key takeaways

• Trust determines AI success, and organisations must prioritise transparency, privacy and clear communication to maintain it.
• Governance enables safe scaling by providing structure, accountability and confidence rather than slowing progress.
• The AI dividend is realised through focused, practical use cases that enhance human decision making and deliver measurable outcomes.