For governments, AI adoption is a skill, not an inevitability.

By Leonardo Quattrucci , 19 May 2025

Leonardo is an Adjunct Professor of Innovation at Sciences Po, a Senior Fellow at the Center for Future Generations and Senior Advisor to Apolitical. Previously, he led technical programmes at Amazon Web Services' Centre for Quantum Networking and served as an advisor to the Presidency of the European Commission.

The most productive firms are the best at integrating innovation into how they work. The same is true for countries. How much, how well and how fast technologies spread across sectors determines which countries are most productive. Why does the European Union’s productivity lag behind the United States? Because of a shortage in the diffusion of digital technologies. 

Technological competitiveness matters for governments too. Advances in artificial intelligence mean that new AI models are accessible to everyone, requiring only basic conversational skills to bring benefits to someone’s work. So they are spreading within the public sector, making competition between states not only about industrial policy, but also about institutional inventiveness and productivity. For example, the UAE announced that it will pilot AI in lawmaking, which could help make policymaking faster, more adaptive and responsive, notwithstanding the risks.      

Because AI develops exponentially, countries that lag in adoption risk creating an irreversible gap with their peers. Not just vis-à-vis firms that produce technology, but also compared to governments that use it to improve – even transform – their traditional functions.

How governments can build AI capability

The AI Readiness Check is an interactive assessment tool developed by Apolitical. Thousands of public servants from dozens of countries are using it, providing a unique insight to the state of government AI readiness. Early insights from the data show there is growing demand to move faster on adoption but real barriers remain.

Tech talent is the biggest gap. Only 17% of respondents in leadership positions attest to plans to develop or recruit AI talent in their organisations. That is the opposite of what happens in the private sector, where companies are shifting investments to hiring AI practitioners. There are isolated exceptions, such as AI safety institutes, but transformation cannot be an exercise in building islands. 

Short of hiring new talent, most governments do not systematically invest in making civil servants more skilled either. Only 30% of public sector managers who support the delivery of AI projects say they have a learning plan to grow AI competencies. Declared ambitions, such as the UK’s plan to double the rate of digital professionals across government, are not turning into execution, yet.

Better management of AI goes hand in hand with better procurement of the technology. Today, if procurement were a piece of software, it would be largely obsolete. Only 17% of leaders say they have a plan for AI procurement. The consequence of this is that governments don’t have the right tools, nor do they have the deep experience to make informed decisions about the technology.

Moreover, just 1 in 3 of public sector leaders using the AI Readiness Check say they have a framework for prioritising investment in potential AI use cases. And less than half of practitioners can explain different approaches to AI procurement. Without measures of success, it is hard to judge what returns on public investments are worth seeking. To make the most of AI innovation, governments must fix procurement.

On this basis, AI adoption in the public sector will not be an inevitability. But it’s an urgent and growing necessity. Building an AI-ready government is a skill and investing in people is a central pillar. Without the freedom to hire “builders,” upskill their workforce, and move the right people in the jobs that fit their strengths, governments’ AI-illiteracy may become an irrecoverable gap. It is skilled personnel that makes good decisions, designs intelligent pilots and procures what the public and the data demand. It’s time public institutions take their own medicine.

Share your thoughts on Leo's piece in the AI in Government Community with more than 8,000 public servants and policymakers from around the world.