Six digital experiments that were never meant to scale

Governments are under pressure to move on digital transformation and AI. But many don’t know where to start – and honestly neither do we much of the time. And that’s OK. Landscapes are shifting quickly, institutional capacities vary enormously, and the gap between a promising pilot and something that actually works at scale is wider than it looks.

Over the past year, the UNDP Istanbul Regional Hub worked with six Country Offices to try something a little different. Through the Digital Transformation Acceleration Facility (DTAF), each received USD 30,000 and technical support to run small, early-stage experiments testing digital ideas. These “Trojan Mice” experiments were low-cost tests designed to surface actual conditions before resources get locked into solutions that may not fit. 

Here’s what we found:

The experiments

Colleagues in Albania tested ways to build AI literacy inside municipalities. Through focus groups in three cities, the team discovered that officials were already using AI tools informally but often without fully understanding the related risk and safety issues. The team developed a simple curriculum mixing demonstrations with hands-on prompting, helping staff to move from tentative experimentation to more confident, deliberate use.

The team in Kazakhstan explored how to strengthen the country’s digital health system. They assessed two public health laboratories used for diagnostics and testing, looking at how day-to-day operations could be improved. Based on this, they developed business cases for IoT sensors and digital twins, which could help monitor equipment performance, environmental conditions and energy use in real time, making laboratories more reliable and easier to maintain. They also brought Ministry of Health officials to Copenhagen to see open-source logistics systems in action.

The Country Office in Montenegro partnered with the city of Podgorica to prototype three small tools: an AR feature for tourist sites, an AR visualization for spatial planning, and an AI-powered chatbot designed to help residents and visitors access basic city information and services. The technology was simple, but it sparked wider conversations inside the city about how emerging tools could support municipal services, with clear potential to be upgraded and expanded with additional functionalities over time.

The team in Serbia mapped the country’s fast-growing but loosely coordinated AI ecosystem. Working with local researchers, the team built an interactive map of actors, strategies, and projects, and drafted priorities for a National Language Technology Programme. The goal was to create a shared picture of the ecosystem that could support more coordinated policy decisions, clearer investment priorities, and future collaboration across government, academia and the private sector.

Colleagues in Uzbekistan designed a navigation mobile app to improve mobility for people with disabilities. Testing with users with disabilities revealed practical constraints: gaps in open-source maps, challenges developing in the local language, and the technical limits of what a pilot can achieve in a short amount of time.

The Türkiye Country Office convened a multi-stakeholder AI workshop during institutional restructuring, engaging key ministries, agencies, academia, and the private sector. The discussions sparked concrete interest in applying AI to government grant evaluations and informed the national AI strategy process. Building on this momentum, UNDP Türkiye supported targeted AI trainings for development agencies and municipalities, helping build trust, strengthen capacities, and set the foundation for follow-on pilots.

What these experiments revealed

Governments often approach digital transformation as a set of discrete tasks: an app, a dashboard, an AI pilot. But these requests frequently mask deeper challenges – coordination failures between institutions, missing data, and accountability gaps.

DTAF was designed to surface what’s actually going on before committing to large-scale solutions. In Albania, the AI literacy workshops revealed how uneven internal processes and skills were across municipalities. The Uzbekistan team’s navigation app exposed gaps in open-source mapping data. The colleagues in Kazakhstan discovered how fragmented the health data ecosystem was only after they began mapping it. In each case, the DTAF experiment worked more like a diagnostic, showing teams the structural challenges that any digital layer would ultimately rest upon.

The work is political, not just technical

Several teams had to adjust their projects not because the technology changed, but because the institutional ground shifted. Serbia reoriented its approach amid political changes. Türkiye slowed down when national AI structures were reorganized. Montenegro’s prototype depended heavily on one overstretched person inside the city administration.

Digital transformation ultimately deals with questions of authority, resource allocation, mandates, and legitimacy. Every decision about data flows, procurement, or system ownership reveals political sensitivities that the use of technology alone cannot resolve. Teams that made progress did so by building trust with shifting actors and staying anchored in the practical public value of their experiments.

Relationships moved work forward when structures couldn’t

Across countries, the biggest enabler was relationships. The progress with the Ministry of Health in Kazakhstan moved well because trust and collaboration were already established. Serbia’s ecosystem mapping came together because organizations voluntarily shared information. The AI workshop in Türkiye created enough cross-institutional trust that new ideas emerged organically. In Uzbekistan the team learned that their app could only sustain if a national body agreed to maintain it. 

What this changes

These projects were small by design. Their value was in creating clarity: helping teams see where data actually sits, which actors matter, where the real bottlenecks are, and which ideas are worth developing further.

That clarity is now shaping where Country Offices invest time, who they partner with, and how they frame digital work with governments. It’s also a reminder that digital transformation is a continuous process of testing, learning, and working with political realities – one that demands patience, relationship-building, and a willingness to let small experiments reveal what large plans sometimes obscure.

For teams starting similar work: don’t mistake the quick ask for the real question. An app, a dashboard, an AI pilot. These are entry points, not endpoints. Use them to get inside the system, to understand its dynamics, to find where energy and momentum already exist. That’s where the real work begins.

Photo by Conny Schneider on Unsplash