Ambition Surges, But Public Sector AI Stalls at Pilot Stage
Across the world, government AI deployment is shifting from experimentation to ambition. KPMG’s Global Tech Report on government and public sector organisations finds that 48 per cent plan to deploy AI use cases into production at scale within the next 12 months. Yet 43 per cent admit they face serious hurdles moving beyond AI pilot projects, revealing a widening gap between strategy decks and operational reality. Leaders report that they are still in the early stages of public sector AI adoption and that challenges around scaling are now clearly visible in the data. Much of the technology budget is still consumed by maintaining legacy systems, slowing the pivot to true digital transformation. As a result, agencies are under pressure to demonstrate quick wins with AI while wrestling with old infrastructure and cautious risk cultures that were never designed for fast‑moving, data‑hungry tools.
Why Scaling Government AI Is So Hard: Data, People, Procurement, Security
The biggest obstacles to government AI deployment are structural, not just technical. KPMG highlights fragmented data and siloed thinking, with agencies struggling to integrate information across functions, departments and broader ecosystems. Only 37 per cent report high maturity in enterprise data systems, even though clean, reliable, integrated and accessible data is essential for AI to add value. An AI talent shortage compounds the problem: governments find it difficult to build the specialised skills needed to design, manage and govern AI systems, and 86 per cent believe that managing AI agents will become a critical capability within five years. Procurement inertia also slows progress, as traditional buying processes are poorly suited to agile, experimental AI projects. Meanwhile, AI cybersecurity risks loom large. Just 38 per cent of respondents consider their cybersecurity capabilities mature, even as 64 per cent plan to increase investment in this area.
Labs, Ecosystems and Partnerships: How Governments Are Trying to Catch Up
To break out of the pilot trap, governments are leaning on public‑private partnerships and new innovation ecosystems. KPMG notes that about 75 per cent of public sector leaders plan to expand their technology partnerships, even as they worry about data protection, security, costs and strategic alignment. One visible approach is the creation of AI labs and innovation centres where vendors, agencies and researchers can co‑develop use cases, share infrastructure and build skills. A recent example is Huawei’s AI Lab and Innovation Centre launched at The Exchange 106, designed to serve the wider Asia Pacific region. The facility showcases AI applications for government and education scenarios, supports digital and energy transformation use cases, and aims to train more local AI talent through closer engagement between industry, research and education stakeholders. Such hubs signal a shift toward outcome‑focused collaboration rather than isolated, one‑off pilot projects.
What the Implementation Gap Means for Vendors and Startups
For major AI vendors, cloud providers and startups, the gap between public sector AI ambition and execution is both an opportunity and a warning. On one hand, governments clearly intend to scale AI, creating long‑term demand for platforms, infrastructure, data tools and managed services. The push to expand ecosystems and partnerships suggests vendors that can navigate compliance, security and multi‑stakeholder governance stand to gain. On the other hand, slow data integration, AI talent shortages and heightened AI cybersecurity risks can stall or shrink deals, especially when projects cannot progress past the pilot phase. Vendors must therefore offer more than generic AI tools: they need end‑to‑end support for migration from legacy systems, robust cybersecurity capabilities, and practical enablement for civil servants. Local startups may find niches in specialised use cases or domain‑specific data solutions, but they must be ready for rigorous security and procurement scrutiny.
Citizens at the Centre: Better Services, But Also New Risks
Ultimately, public sector AI is not about algorithms; it is about citizens’ daily experiences. KPMG observes that leading governments are pivoting from process‑centric to citizen‑ and outcome‑centric approaches, using AI to improve service delivery, reduce wait times and detect fraud more effectively. Integrated data and well‑governed AI could help agencies anticipate needs, personalise interactions and allocate resources more fairly. Yet the same technologies raise serious concerns. Poor data quality or biased models can entrench discrimination, while opaque decision‑making undermines trust. Expanding data sharing to power AI can increase surveillance risks if safeguards are weak, and immature cybersecurity capabilities heighten the danger of sensitive information being exposed or manipulated. To earn public confidence, governments will need transparent governance frameworks, strong data protection, clear channels for redress and genuine public engagement on how AI is designed and deployed in essential services.