Why public administration and GovTech are struggling to become a true ecosystem

Two-thirds of Italian public administrations (PAs) have experimented with artificial intelligence, yet their ability to adopt it in a structured way remains limited. As the National Recovery and Resilience Plan (PNRR) approaches its conclusion, Italian public administrations have pilot projects, unspent funding, and at last, a well-defined regulatory framework. But what’s still missing is the capacity to turn a funded pilot into a stable public service.

Alongside a public sector that continues to struggle to adopt AI in a mature way, there is a growing GovTech industry that has yet to connect with public administrations at a scale that would make this technology sustainable. The bottleneck lies precisely in this disconnect between administrations that lack the capacity to procure and integrate innovative solutions, and companies that can’t secure enough public contracts to grow. In a report produced by SDA Bocconi as part of AI-PACT (Artificial Intelligence for Public Administrations Connected), the European Digital Innovation Hub coordinated and led by SDA Bocconi, this dual weakness is substantiated for the first time.

An initiative recognized by the European Commission

AI-PACT, which received the European Commission's Seal of Excellence, was created to move beyond the anecdotal approach that has characterized discussions of AI adoption in the Italian public sector to date. The research was conducted by colleagues Tommaso Bianchi, Maria Cucciniello , Niccolò Cusumano , Leonardo De Rossi , Lisa Fontanella , Giulio Guidotti , Giulia Piras , Gianluca Salviotti , Aristea Saputo , and Roberto Vivona . The team focused on three questions: How prepared are public administrations to adopt AI as a service technology? Which organizational capabilities are already in place, and which are still missing? And what is the current status of the GovTech ecosystem, which is expected to meet public-sector demand for innovation?

To answer these questions, researchers had to adopt a balanced perspective on both sides of the innovation ecosystem. As far as existing maturity models (tools used to assess an organization's readiness to adopt a given technology), these were developed for the private sector and as such are poorly suited to organizations whose mission is to create public value through accountability, fairness, and democratic legitimacy. So a diagnostic tool specifically designed for the public sector was needed, one calibrated to the risk-based compliance requirements of the EU’s AI Act (the world’s first comprehensive AI law).

A framework that supports the entire innovation lifecycle

AI-PACT is designed as a customer journey that guides public administrations throughout the entire innovation lifecycle, from the initial assessment of organizational readiness all the way to impact measurement, following the theoretical framework of innovation diffusion. This approach makes the findings of the study particularly useful for decision-makers because each stage provides practical guidance on where action is needed.

The first tool is the AI-PACT Maturity Model, which builds on the Technology-Organization-Environment framework to assess organizational readiness across six dimensions. External Context covers political stability, citizen engagement, and communication of initiatives. Management and Organization measures organizational culture, training, and strategic vision. Governance revolves around procurement, compliance, and the clarity of roles and responsibilities. Human Resources evaluates internal skills, partnerships, and the ability to attract talent. Technological Resources examines data accessibility and the robustness of infrastructure. Finally, Value assesses both the perceived benefits of AI and the organization's ability to evaluate outcomes before and after adoption.

The following stages shift the focus to practical experience.

• The Sustainability Assessment examined the long-term viability of AI pilot projects. It found that most initiatives were launched as one-off experiments without a permanent funding model or an internal governance structure capable of taking them beyond the pilot phase. In most cases, sustainability was more of an aspiration than a design principle.
• The Impact Assessment then explored how users experienced the pilot solutions, showing that perceived value depends far more on the quality of organizational support than on the characteristics of the technology itself.
• The final stage examines the supply side through the Startup Mapping Tool, which measures how prepared Italian GovTech companies are to serve the public sector across three strategic dimensions.

Where public administrations are strong—and where they still fall short

The Maturity Model assesses organizational readiness on a scale from 1 to 5. Scores below 2.5 indicate that capabilities are fragile, scores up to 3.5 signal that they are present but not yet systematic, and higher scores reflect mature capabilities. Italian public administrations achieve an average score of 2.79, placing them in the intermediate range, where there is awareness but it has not yet translated into systematic practice. But this average conceals a high degree of polarization: Only 12% of organizations reach the highest level, while 27% linger in the lowest, with differences driven more by organizational size than by geographic location.

The strongest dimensions are External Context and Management Culture. This is where the PNRR has had its biggest impact, accelerating an innovation-oriented culture more than shoring up the capacity to turn innovation into permanent public services.

By contrast, performance is poorest in Governance and Human Resources, which are also the two dimensions that matter most for long-term sustainability. Governance is the least mature dimension overall, with a score of just 2.36 on a scale where 2.5 marks the threshold for adequate performance. Procurement remains the weakest link because most acquisitions still rely on general ICT procurement procedures rather than processes designed specifically for AI solutions. Compliance with European requirements is only beginning, and specialized roles that should oversee it, such as AI Officers and Data Stewards, are almost entirely absent. As for Human Resources, the main limitation lies in the lack of stable partnerships with research institutions and the private-sector innovation ecosystem. Smaller public administrations collaborate far less with universities, research centers, and innovative companies, leaving them with no access to cutting-edge expertise, which would be impossible for them to develop internally. This is compounded by a heavy reliance on external vendors, which in non-urban municipalities accounts for as much as 62% of technological infrastructure. As a result, the ability to steer and oversee innovation increasingly resides outside the administration itself.

Strategic priorities

Taken together, the findings point to a conclusion that applies both to public-sector leaders and to policymakers responsible for shaping the broader system. The return on investment in AI depends on the organizational capacity to integrate tools, skills, and partnerships in ways that allow innovation to survive beyond the period of extraordinary funding.

The most urgent priority is Governance, the weakest dimension identified in the assessment. Building adoption capacity means establishing permanent specialist roles, such as AI Officers and Data Stewards, as well as procurement procedures designed specifically for AI, not adapted from traditional ICT purchasing.

The second priority concerns Human Resources, the other critical deficiency. The objective is to reduce dependence on external vendors by investing in internal capabilities and long-term partnerships with research organizations and the broader innovation ecosystem.

The third priority addresses the inadequacy of the Value dimension and focuses on evaluation. Without structured processes to measure impact before and after adoption, every pilot project remains an isolated initiative, unable to generate learning that can benefit future implementations.

Ultimately, the challenge for the system today is to bring together the two disconnected sides by making public-sector demand better positioned to adopt innovation, and the GovTech on the supply side better equipped to respond.

AI e PA: stato dell’arte, priorità e traiettorie di sviluppo . Rapporto di ricerca. (AI and the PA: current status, priorities, and development pathways. Research report, published in Italian)