While the transformative potential of AI for public service delivery is undeniable, the path from policy ambition to practical implementation is fraught with complex challenges. There are government departments, and even parts within those organisations, at both ends of the spectrum with relation to appetite for AI adoption – eagerness to move forward as well as hesitation, potentially due to the shadow of Robodebt.
As Leidos Australia Vice President Science Engineering & Technology, Murray Bruce said, “Even though Robodebt wasn’t an AI program, that idea of what trust you place in digital capabilities and how that is used to transform the way you deliver services or outcomes is at the top of mind for these agencies.”
This reticence by agencies on one hand to adopt AI for fear that they would tread on similar issues leads to a level of caution. While on the other hand, AI is already supporting government operations through everyday SaaS platforms, and forward-thinking agencies champion targeted adoption in mission-critical areas.
Navigating the Misconception Minefield
Historically, some people might have believed that there was a one size fits all approach to AI, but today, we’re seeing a wide diversity of AI tools – they are specialised, the benefits they deliver in their niches are compelling and it’s about finding the right tool for the right job. “A common myth was that there can be a single AI approach to everything,” Bruce said. “Some would say, well, you know, we can use ChatGPT to do all of this, and consequently they’re not thinking about the full AI toolkit.”
In addition, success relies not only upon selecting the right tools but requires consideration of user interface design, data engineering, workflow optimisation, and traditional computing capabilities alongside AI solutions.
Another misconception surrounding AI adoption is that it’s easy. The reality is far more complex, particularly when vendors attempt to force-fit their existing AI capabilities rather than selecting the right tool for each specific use case.
Meanwhile, AI is naturally embedding itself into government operations through commercial platforms. “You don’t really have control over the infusion of AI into government, particularly through your off-the-shelf SaaS services,” Bruce said. With Microsoft 365 delivering Copilot, Salesforce integrating AI features, and similar developments across enterprise software, agencies are experiencing AI adoption whether they’re actively pursuing it or not.
Building for Success: Infrastructure and Strategy
The key to sustainable AI implementation lies in moving beyond ad-hoc project approaches toward systematic capability building. This requires what industry professionals are calling an “AI factory” approach, and creating repeatable, scalable capabilities rather than one-off solutions.
“We’re investing in the AI factory, which is the partner capability to the software factory tooling, to enable us to ask the full AI lifecycle effectively,” Bruce said. “This approach shifts focus from individual projects trying to ramp up capabilities to do things effectively toward having a consistent capability, which we can apply across numerous use cases.”
Critical to this approach is disciplined use case selection. Successful AI adoption requires selecting the right sort of use cases for which the AI technology as it exists today, is suitable to address. This is important to get right, as the alternative is burning through social licence with stakeholders by attempting to solve overly complex problems that current technology cannot adequately handle – this is, ultimately, what the high failure rate of AI projects comes down to.
Data sovereignty requirements add another layer of complexity but provide clear qualification criteria. “The first question you need to do to qualify a capability is, is that capability hosted onshore?” Bruce said. This requirement, combined with IRAP assessment needs, creates a systematic filtering process for vendor selection. While Leidos-delivered solutions are designed to have data sovereignty to start with, many AI solutions today face similar hurdles to those seen when governments first began adopting cloud technologies.
The Path Forward: Sustainable Implementation
The most promising applications for government AI lie in mission-critical areas where human operators face overwhelming data volumes. There, the value proposition is clear: “How do we enable those analysts or operators or team members to focus on their tradecraft and bring them the high-value data rather than bring them all the data?” Bruce said.
Success in this environment requires embracing a framework built around speed, scale and security to deliver the mission and value. This means moving away from the handful of data scientists in the corner and toward democratising AI capabilities across entire development workforces.
The path forward requires acknowledging that AI implementation is fundamentally about digital transformation, not just technology adoption. Agencies that recognise this distinction, invest in proper infrastructure, and maintain focus on mission value rather than technological novelty will be best positioned to harness AI’s transformative potential while maintaining the public trust essential to democratic governance.
With over 25 years of local experience, Leidos provides IT services to federal government agencies to support service delivery, digital modernisation and the uptake of emerging technologies – securing, modernising and sustaining Australia’s most critical systems. Explore more at Digital Modernisation | Leidos
Source link
