A new framework outlines how governments can scale agentic AI to improve efficiency and deliver public value. With a US$9.8 trillion opportunity by 2034 and 90% of agencies exploring adoption, success hinges on governance, risk control, and prioritizing high-readiness use cases.

 

The adoption of Agentic AI could unlock a US$9.8 trillion opportunity for public sector digital transformation by 2034, reads a new report by the World Economic Forum (WEF), Capgemini, and the Global Government Technology Centre Berlin. The report aims to transition agentic AI from experimentation to scalable public value. 

“Governments that approach agentic AI with discipline and strategic intent will not only improve their own operations, but they will shape the norms and expectations for how this technology is governed in the public interest,” says Martina Klement, Berlin’s Permanent Secretary for Digital Transformation and Administrative Modernization and Chief Digital Officer.

During the last few years, governments have invested heavily in digital transformation to move paper processes to screens. However, demographic decline and tightening fiscal spaces now require administrations to do more with fewer people. 

Traditional digitalization preserved the underlying logic of public administration, but agentic AI marks a shift because systems can plan, decide, and act across multi-step workflows. Unlike generative AI, which primarily creates content, agentic AI executes actions and decisions to reach complex goals.

The Making Agentic AI Work for Government: A Readiness Framework argues that this technological evolution could unlock a US$9.8 trillion opportunity for governments. The report indicates that 90% of institutions plan to explore or deploy agentic AI within two to three years. For this transformation to succeed without eroding transparency or accountability, governments must implement strategic sequencing and governance.

Defining the Agentic Opportunity

Agentic AI systems coordinate processes, integrate information from multiple sources, and adapt actions based on evolving conditions. These systems operate under human oversight, either as human-in-the-loop for key decisions or human-on-the-loop for supervisory control. 

This technology allows public servants to focus on meaningful work that requires human judgment, such as policy interpretation and direct engagement with citizens.

The report introduces a framework with four building blocks to guide decision-making. It assesses functions against potential and complexity, providing a clear picture of where governments can act with confidence. 

This systematic approach is essential because Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027 due to escalating costs or inadequate risk controls.

A Function-Based Assessment Lens

The framework evaluates government activities by function rather than department. A function is a recurring workflow that delivers a clear result, such as eligibility assessment or identity verification. This lens reflects operational reality and promotes cross-domain learning. The research identifies 70 core functions across nine categories, including public services, policy planning, and IT infrastructure.

A Roadmap for Local Implementation

The report provides a baseline, but argues that feasibility depends on local conditions. Governments must translate the topography into a national roadmap through six steps:

Establish local baselines: Assess digital maturity, data infrastructure, and workforce capabilities.
Develop risk-mitigation strategies: Address sustainability, workforce readiness, and infrastructure resilience.
Reassess function-level scores: Use domestic knowledge to adjust global estimates.
Sequence implementation: Start with high-readiness functions to build organizational confidence.
Validate through testing: Run small-scale pilots with real users and live data.
Iterate as conditions evolve: Regularly revisit assessments as technology and regulations change.

Case Studies in Agentic Deployment

Early implementations show that agentic AI creates tangible value in government operations. In Ukraine, the Diia.AI national assistant enables citizens to access certificates through natural language. Since September 2025, more than 290,000 citizens have used the system, which issued over 7,000 certificates.

In Germany, the Federal Ministry for Digital Transformation and Government Modernization developed an AI-based construction permit system. The system screens applications, maps them against legal norms, and generates drafts for permit decisions. This targeted approach allows the state to apply AI capacity only where it adds value.

In the United Arab Emirates, the Federal Authority for Government Human Resources launched an HR AI Agent. The system autonomously resolves over 80% of legislation and policy inquiries for 50,000 federal employees. By automating routine transactions, HR professionals can shift toward high-value activities like performance optimization.

The German Federal Employment Agency also piloted an agentic solution to translate requests for change into Jira tickets. The pilot suggests this solution could save about 150 hours a month for the development team.