A recent global survey of nearly 500 senior government executives highlights a challenge that cannot be ignored. While enthusiasm for artificial intelligence (AI) and generative AI (Gen-AI) is high, adoption remains limited. Only 26 per cent of government organisations report integrating AI at scale, and just 12 per cent have Gen-AI solutions in operation.

This gap between ambition and reality matters. AI has the potential to transform service delivery, reduce costs, and improve citizen outcomes. Without careful planning and strong foundations, however, governments risk falling into cycles of stalled pilots, unrealised benefits, wasted investment, and declining public trust.

Why the Gap Exists

Several factors are slowing adoption. Legacy systems and technical debt consume resources that could be directed to new initiatives. In many agencies, critical infrastructure is outdated, which forces teams to focus on maintenance rather than transformation (Read more here).

Governance frameworks are another obstacle. The EY survey found that 62 per cent of leaders view data privacy and security as significant barriers to AI success. More than half lack a clear data transformation strategy. Without a coherent plan, adoption efforts are fragmented, reactive, and vulnerable to compliance risks.

Culture and capability also play a role. A small group of pioneers have built robust digital foundations and digitised their core processes. The majority remain in the early stages of experimentation. This creates a widening gap between leaders and followers, which risks uneven service delivery and a lack of resilience across the sector.

The Risks of Inaction

If these issues remain unresolved, governments face serious consequences. Technical debt will continue to grow, making future modernisation even more costly. Citizens, accustomed to seamless digital services in other parts of their lives, will grow dissatisfied with slow or inconsistent government systems.

There are also risks to legitimacy. AI can bring efficiency, but if automated decisions lack transparency or cannot be contested, public confidence will decline. Trust in algorithms is central to the success of any government system. Once lost, it is extremely difficult to restore, as proven in the case of the Robodebt or Horizon scandals. (That is why Castlepoint created truly Explainable AI. To ensure every AI-supported decision is transparent, trustworthy, and accounted for.)

Closing the Gap

Agencies need to move decisively if they want to close the adoption gap. Modernising legacy systems with effective AI should be treated as a priority, not an optional expense. AI programs must be grounded in governance frameworks that protect privacy and national security, embed transparency, and ensure accountability.

The benefits can be significant, and completely transformative. Even the most sensitive and highly governed organisations have been able to move forward with adopting AI at scale: for example, the UK Ministry of Defence, who are deploying autoclassification across the estate to remove human error from the security labelling process.

Many other government entities have also achieved significant benefits, while managing the risks, using Explainable, sustainable, and accurate AI:

·       The Commonwealth Treasury has been using AI for records management since 2019, relating items across the environment into Virtual Records, and transforming their Harradine reporting and strategic governance

·       Multiple Departments have used AI following a MOG change, to identify records for transfer or disposal based on current and legacy Functions

·       The Australian Fisheries Management Authority was able to successfully decommission their file share and legacy EDRMS, instead adopting manage-in-place AI compliance

·       State and Federal agencies have used AI to find evidence of child abuse and assault across massive data sets, exposing historical abuse that was not discovered using traditional ESI processes

·       A State Government department transformed their manual, high-effort Information Asset Register to an automated, streamlined capability using AI autoclassification.

Technology is a major part of the solution. AI that does not introduce technical debt, complexity, user impacts, or burden the governance team is not just achievable – it's proven. But there are governance changes needed as well.

Workforce capability requires investment. Digital literacy should be developed across all levels of government, not just within technical teams. Building a workforce that understands the opportunities and risks of AI will create the confidence needed to adopt at scale.

Equally important is the need to build public trust. Citizens must understand how AI is being applied, how decisions are made, and how their information is protected. Without transparency and accountability in the AI that is deployed, adoption will remain constrained by scepticism.

A Path Forward

The gap between ambition and reality is not insurmountable. Governments that treat AI as a strategic capability, backed by strong foundations and citizen trust, will be able to move from pilots to transformation. Those that do not risk locking themselves into cycles of inefficiency, stakeholder dissatisfaction, and potentially even censure.

Castlepoint Systems was founded to help address these very challenges. Our platform was designed to modernise data governance, strengthen compliance, and reduce risk, creating the foundation that agencies need to adopt AI with confidence. You can read more about our work in government here, and speak to us any time about the challenges and opportunities of accurate, scalable autoclassification.

‍