How Australian government agencies can get more from AI through smaller bets than big programs
In this ADAPT Insider episode, ADAPT Executive Advisor, former Chief Digital Innovation Officer and BCG Partner David Heacock explains why government should stay cautious on AI, build more capability internally, and focus on smaller, lower-risk delivery.Government is under pressure to do more with less, and AI is now part of that conversation. That does not mean moving fast.
In government, the cost of getting AI wrong is higher, which makes caution a strength rather than a weakness.
In this ADAPT Insider conversation with Anthony Saba, Managing Director – Transformation Services at ADAPT, David Heacock, ADAPT Executive Advisor, former Chief Digital Innovation Officer and BCG Partner, argues that public sector leaders are right to take a measured approach.
The gap between vendor hype and real-world performance is still too wide, and in government that gap can damage trust, disrupt services, and create harm.
Listen to the full episode on Apple Podcasts, Spotify, or YouTube.
Key takeaways:
- Move carefully where the risk of AI failure is high.
- Treat governance and decision-making as the main delivery challenge, not the technology alone.
- Break large programs into smaller, outcome-driven pieces and build more capability internally.
Government is right to be cautious on AI
He argues that AI should be introduced carefully in public sector settings.
Unlike traditional systems, generative AI is probabilistic, which means it does not always produce the same answer in the same way.
That creates a different risk profile, especially in government environments where decisions can affect people’s lives.
This is why a slower pace makes sense. Government does not need to lead the market on AI adoption.
It needs to understand where the tools are useful, where precision matters, and where experimentation is safe.
The bigger problem is organisational friction
He is clear that technology is no longer the main constraint. The harder problem sits in governance, decision-making, and organisational design.
Faster tools do not remove those bottlenecks. They hit them sooner.
He points to executive choke points, where delivery teams move quickly but decisions still stall further up the chain.
He also warns that uncertainty often leads to more governance, more scrutiny, and more delay. In practice, that usually makes delivery slower rather than safer.
Smaller internal bets will work better than large programs
He also argues for a different delivery model. Rather than packaging large problems into single vendor-led programs, agencies should break work into smaller pieces with tighter accountability and clearer outcomes.
He sees strong potential in building more capability internally and focusing on use cases that sit between personal productivity tools and large enterprise systems.
Areas like policy analysis, legislative synthesis, and information-heavy work are a better fit for small, targeted AI projects that can prove value early without creating unnecessary risk.