From experimentation to public value: Why governments need to learn together

Governments today are operating in a period of rapid and profound change. Technological advances, increasingly complex policy challenges, fiscal pressures and rising citizen expectations place growing demands on public institutions. Yet, many governments are struggling to adapt, reform and deliver at the required pace and scale. In response, governments are increasingly exploring new ways to improve how they design policies, deliver services and create public value. Artificial intelligence (AI) and other emerging technologies are important tools in this effort, offering new opportunities to enhance public services, strengthen internal operations and support better decision making.

However, previous years of technological adoption have shown that it alone is not the answer. Many governments already have examples of AI experimentation. What matters now is whether these efforts can translate into sustainable improvements in public services and government performance.

To help address these challenges, the OECD, together with the European Commission and a growing group of practitioners from across governments are launching the Gov2Gov Community of Practice on Innovation Incubation (AI and Emerging Technologies), a space for public servants to learn from one another’s experiences, share practical approaches, and collectively explore how governments can turn experimentation into public value.

AI as a test case for government experimentation

Governments across the world are increasingly adopting artificial intelligence (AI) and other emerging digital technologies in an effort to respond more effectively to increasingly complex challenges while improving public services and government performance. The OECD’s Governing with Artificial Intelligence report analysed 200 real-world examples of AI use across 11 core government functions, including public service delivery, justice, fraud detection, public finance and civil service reform. The report found that governments are increasingly experimenting with and adopting AI, but that many continue to face practical challenges related to governance, capabilities, implementation and scaling.

Therefore, experimentation is only the beginning. Many governments continue to face a common challenge: how to move from exploration, pilots and isolated use cases to responsible adoption, sustained implementation and scaling. To do it, governments need to learn how to turn experimentation into learning, learning into better decisions, and better decisions into public value.

Recent OECD work on Generative AI experimentation in government reinforces this point. It finds that public servants are increasingly adopting generative AI, often ahead of formal governance arrangements, and argues that structured experimentation is a critical pathway for governments to build capabilities, manage risks and support responsible technology adoption before full-scale deployment. It also highlights that governments remain at an early stage of learning how to experiment systematically with generative AI with monitoring and evaluation also identified as critical weaknesses.

Successful AI adoption in government rarely depends on technology alone. It depends on whether public institutions have the organisational conditions, experimentation practices, governance arrangements, skills and learning mechanisms needed to test, assess, adapt and scale new approaches responsibly. As governments navigate these challenges, many are finding that no single administration has all the answers. Learning from others’ experimentation and incubation efforts can help governments avoid repeating mistakes, adapt promising approaches and accelerate learning.

From pilots to public value

Governments do not simply need more pilots. They need stronger processes for turning pilots into public value: better outcomes for people and society that respond to public needs while sustaining trust, legitimacy, fairness and accountability. Experimentation can help governments understand what creates this value before a solution is institutionalised or scaled, by testing assumptions, generating evidence and learning what works, what does not and under which conditions, an approach reflected in the Declaration on Public Sector Innovation. Today, many governments are grappling with similar questions around adoption, governance and scaling, while looking for practical ways to make experimentation more systematic and useful.

A new OECD working paper on incubators and accelerators for public sector innovation shows that governments are under increasing pressure to develop, test and scale solutions across policies and services, while navigating fiscal constraints, administrative complexity and compliance-focused cultures. The paper analyses 128 cases of public sector incubators and accelerators from 51 countries and six international organisations, showing how these mechanisms can help governments experiment safely, act more quickly and embed more adaptive ways of working.

The same paper also points out that incubators and accelerators cannot simply be copied from the private sector into government. They need to be adapted to public sector realities, including risk aversion, accountability requirements, fragmented responsibilities, political cycles and resource constraints. They must also operate within a broader public mission, where considerations such as public value, legitimacy and trust play a central role in shaping how innovation is designed, implemented and assessed.

This lesson is especially relevant for AI and emerging digital technologies. Where their use may affect public services, fundamental rights, privacy and data protection, experimentation needs to be purposeful, structured and connected to wider governance arrangements. This means testing not only whether the technology works, but also whether appropriate and effective guardrails are in place. Without this, pilots may remain isolated exercises rather than sources of learning for broader transformation.

This is also increasingly a question of public trust. Recent OECD findings on trust in public institutions suggest that people are generally more optimistic about AI’s potential to improve services and efficiency than about governments’ ability to ensure transparency, fairness, privacy protection and human oversight when deploying these technologies. Structured experimentation can therefore serve a dual purpose: helping governments understand what works technically while also testing and strengthening the governance arrangements needed to make AI trustworthy in practice. Experimentation can help governments assess not only whether AI works, but also whether its deployment can sustain trust and legitimacy while creating public value.

Experimentation needs structure

Effective experimentation is not improvisation. It requires clear problem definition, prioritisation, governance, evaluation and feedback loops.

The OECD is also developing two companion reports on measuring the impact of AI investments in government. This work responds to a significant gap in current practice. The 2025 Digital Government Index (DGI) shows that only 28% of OECD countries have conducted any financial or non-financial impact measurement of AI use cases in government. Collectively, the reports examine why stronger impact measurement is needed to move from pilots to public value, and how measurement can be embedded across the AI investment lifecycle to inform decisions about whether initiatives should be continued, adapted, scaled or stopped. Together, they reinforce the importance of planning for measurement and evaluation from the outset rather than treating them as an exercise undertaken after deployment.

This work, together with OECD work on generative AI experimentation, points to several actions that are critical for governments: making experimentation frameworks future-fit, ensuring efficiency through clear prioritisation and problem definition, planning experiments and evaluation approaches from the outset, and creating environments that encourage responsible experimentation. Without these conditions, experiments are likely to remain isolated pilots rather than informing broader policymaking on the role of AI in government.

The OECD paper on incubators and accelerators reaches a similar conclusion from a public sector innovation perspective. It identifies five key dimensions that shape how incubation and acceleration programmes can be adapted to government: strategic purpose, ecosystem engagement and co-ordination, organisational context, performance mechanisms and value creation.

Table 1. Key dimensions shaping how incubation and acceleration programmes can be adapted to the public sector

Dimension	Description
Strategic purpose	Incubators and accelerators act under the general objectives of public sector organisations to serve the public good.
Ecosystem engagement	The innovation ecosystem plays a critical role surrounding these programmes: different ecosystem configurations and connections influence the shape of incubators and accelerators.
Organisational context (including role of government)	The arrangements by which a programme is organised in public institutions, communities, and networks.
Performance portfolios	The range of services and support structures that make up the incubator or accelerator programme offering.
Value creation	Assessing and evaluating the value (both direct and indirect) and impact which a programme generates for its participants, the public sector in general, and the broader innovation ecosystem.

Source: OECD (2026), Incubators and accelerators for public sector innovation.

These dimensions are not only relevant for formal incubator or accelerator programmes. They highlight a wider need for governments to build the conditions for experimentation, adoption and scaling: spaces where practitioners can compare approaches, test assumptions, understand risks, and learn from the experience of others.

AI is the current frontier, but the challenge is broader

AI is currently one of the most visible and urgent areas of experimentation in government. But it is not the only one.

A 2025 OECD policy paper on civic participation and emerging technologies shows how governments are also exploring tools such as blockchain and virtual reality, alongside AI, to reduce barriers to participation, strengthen government capacity and support more intelligible and accountable participatory processes. The report also stresses that these technologies come with risks and complexities that require careful experimentation, governance and evaluation.

This suggests a broader lesson. Whether governments are experimenting with AI, blockchain, immersive technologies or other emerging tools, many of the implementation questions are similar:

• What problem are we trying to solve?
• Is this technology appropriate?
• How do we test it safely?
• How do we assess public value?
• How do we govern risks?
• How do we scale what works?

These are not questions that any government can answer alone. While contexts differ, many governments face similar implementation challenges when experimenting with, adopting and scaling emerging technologies. Sharing experiences and lessons across jurisdictions can help accelerate learning and reduce duplication of effort.

Governments can learn faster together

The OECD Gov2Gov Innovation Incubator showed the value of cross-border learning among government teams facing similar innovation challenges. The first phase connected “challenge owners” with “solution providers” from other governments, creating structured opportunities to exchange methods, tools, experiences and lessons.

The assessment of the Gov2Gov Innovation Incubator found improvements in participating teams’ innovation readiness. Challenge owners reported a 38.5% increase in their ability to address barriers to scaling, a 35.6% improvement in their understanding of the solution’s key goals and characteristics, and a 28.6% improvement in confidence in implementing their solutions.

One important insight emerging from these results is that cross-border learning can help public servants see their own challenges differently. It can expose teams to approaches, tools and lessons that would otherwise remain embedded in individual administrations. It can also help governments avoid reinventing solutions to similar problems.

Introducing the Gov2Gov Community of Practice

Building on this experience, the OECD is launching the Gov2Gov Community of Practice on Innovation Incubation: AI and Emerging Technologies.

The Community is not an incubator itself. Rather, it is a practitioner-led space for public servants working on experimentation, innovation, AI, emerging technologies, public service delivery, incubators, labs and digital government. Through regular exchanges, practitioners can explore common implementation challenges and learn from one another’s work.

The Community forms part of the next phase of the Gov2Gov initiative, developed under the OECD’s work programme supported by the European Union’s Horizon Europe Programme. It will explore practical questions such as:

• How do governments know whether an AI or innovation initiative is worth pursuing?
• How can public teams avoid technology-first approaches?
• How can experimentation support responsible AI adoption and effective guardrails?
• What organisational conditions are needed to scale innovation?
• How can governments measure public value beyond efficiency gains?
• How can public innovation labs, incubators and communities sustain learning over time?

Sessions will use real, practical challenges faced by governments as anchors for discussion. These challenges will be selected and refined to ensure that they are relevant, transferable and useful to the wider community. While challenge owners will benefit from the discussion, the primary objective is not to develop solutions for individual challenges. Instead, real implementation questions are used as catalysts to surface practical experiences, tools, methods and lessons that can benefit a wider set of governments.

How this Community can support governments

Helping governments experiment more effectively. Governments are increasingly experimenting with AI and emerging technologies, but many are still learning how to do so systematically and responsibly. Structured experimentation can help governments build capabilities, manage risks and support responsible adoption before full deployment.

Supporting the journey from pilots to scale. Scaling remains a persistent challenge. Moving from pilots to sustained adoption depends on organisational capabilities, governance arrangements, leadership, skills, evaluation and learning mechanisms.

Making practical knowledge easier to share and reuse. Much of the knowledge governments need is practical and tacit. It sits in teams, projects, labs, incubators and public servants’ day-to-day experiences. A Community of Practice can help surface this knowledge, make it shareable, and support governments to adapt it to their own contexts.

What happens next

The Gov2Gov Community of Practice will convene public servants and government practitioners interested in experimentation, AI, emerging technologies, public sector innovation, public service delivery and the role of incubators, labs and collaborative mechanisms in supporting government transformation.

Practitioners interested in joining the Community are invited to express their interest through the registration page.

Participants will be invited to share practical challenges, use cases, tools, methodologies and lessons learned. As the Community evolves, the OECD will continue sharing insights through OPSI channels, including follow-up blogs and reflections from Community sessions.

The ambition is simple: to help governments learn together so that experimentation with AI and emerging technologies can move beyond isolated pilots and contribute to public value.

Do you have an idea, challenge or experience that could benefit the wider community? We’d love to hear from you at [email protected].

About the authors

Emanuele Colini is an IT Executive in the IT and Innovation Department at INPS (Italy), with 20 years of leadership experience across the public and private sectors. He focuses on scaling agile delivery, strengthening IT governance and advancing enterprise AI strategy and adoption to modernise public services.

Gerhard Embacher-Köhle is Principal Consultant and Innovation Manager at the Austrian Federal Computing Center, Public Sector Innovation Specialist at the World Bank, faculty member of the Vienna University of Economics and Business, and startup mentor at the University of Vienna.

Andrea Erdei is a Policy Officer for Innovation in Government at the European Commission’s Directorate-General for Research and Innovation, focusing on public sector innovation, organisational development and participatory approaches to change.

Paul Grignon manages public sector innovation projects at Ecolab, the French government’s innovation lab for the ecological transition. He combines behavioural science, data science and evaluation to support human-centred, evidence-based approaches to the ecological and digital transitions.

Juan Leal is a telecommunications engineer with over 20 years of experience in the Andalusian Regional Government (Spain), spanning digital transformation and public sector innovation. He focuses on building innovation capabilities and supporting the design and implementation of the Public Innovation Strategy.

Ricardo Zapata is a Policy Analyst at the OECD’s Digital Government, Innovation and Public Services Unit, focusing on digital government maturity and AI in the public sector.

Nora O’Donnell is a Policy Analyst at the OECD’s Digital Government, Innovation and Public Services Unit, focusing on public sector innovation and innovation incubators.

This project and blog are funded by the European Union. Its contents are the sole responsibility of the OECD and do not necessarily reflect the views of the European Union.