Co-creation and public AI as the next leap of Digital Public Infrastructure (DPI)

The conversation around digital public infrastructure (DPI) has quickly evolved from technical architecture to one centred on inclusion, trust, and real-world usability. 

DPI refers to interoperable digital systems, namely digital ID, payment platforms and data exchanges, that are open, secure, and designed to deliver both public and private services. 

During Glue of GovTech: Co-Designing Digital Public Infrastructure (DPI) that Reaches the Last Mile panel at GovInsider’s Festival of Innovation (FOI) 2026 event, speakers shared more about how to design interoperable, citizen-centric digital systems that leave no one behind.

“Digitisation isn’t always equal to transformation. You have to talk about those who are excluded,” said Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH’s GovStack Advisor, Puja Raghavan, who set the tone of the discussion as the moderator. 

Access Partnership’s Director of Multilateral Relations, Data Policy and Partnerships, May-Ann Lim, proposed for governments to weave in different levels of co-creation across the development of DPI. 

“The idea of co-creating something with people who don’t quite understand the full entire stack begs the question, what point do you engage?” she reflected. 

Instead of advocating blanket participation, Lim argued that governments must carefully decide where co-creation is most meaningful.   

A more centralised approach might be necessary when it comes to foundational layers such as infrastructure procurement or system architecture. 

Co-design might be more critical at the application layer, where citizens directly interact with services. 

“There are certain places where we want to absolutely say what exactly the needs of the community are. Don’t do it from a top-down perspective,” she added.  

Interoperability is more than just technology   

Sri Lanka’s DPI Technical Advisor, Dasun Hegoda, highlighted that if DPI is to function as a cohesive system, interoperability was the backbone.  

He explained that interoperability in government comprises four interrelated layers.  

The process would begin with legal interoperability, which refers to whether institutions have the legal mandate to share data. Without this foundation, integration cannot proceed.The next layer was organisational interoperability, which would help align workflows across agencies. 

This was followed by semantic interoperability, ensuring that data is understood in the same way by different parties. Without this shared understanding, the exchange of data could lead to confusion.  

Technical interoperability was about having these three layers in place, connected by APIs and protocols. 

“This layered approach reveals why many digital initiatives fail. Usually, we think only about the fourth layer. But you need to have all of these four layers,” he said. 

To address this, Hegoda suggested putting in place foundational principles such as open standards, common data models, and “security by design” and “privacy by design.”   

“You can expose data, but what if exposing data does harm?”, he asked, adding that continuous monitoring is important to ensure that systems evolve safely and effectively over time. 

Leveraging open-source AI 

As artificial intelligence (AI) is increasingly embedded into national systems, governments are grappling with whether to build or buy. 

AI Singapore’s Head of Partnership, Strategic and Growth, Mark Pereira, argued that open-source AI offers a crucial middle ground. 

“If you want to go fast, you buy. If you want to build for the long term for a strategic perspective, you build.” 

But he emphasised that the decision was not binary. Between fully proprietary systems and building from scratch, there lies a “sweet spot” where governments can adapt open-source models to fit global AI for local needs. 

This approach allows countries, especially in Southeast Asia, to maintain relevance in a rapidly evolving AI landscape. 

“We want to make sure that we are moving ahead with the curve but also bringing our own preference and perspective into AI,” he noted. 

According to him, the stakes are particularly high for regions underrepresented in global datasets. He cited data that less than one per cent of open web data reflects Southeast Asian contexts, creating a structural imbalance in how AI systems understand culture, language, and society. 

In this context, open source is not just a technical preference. It is a strategic tool for digital sovereignty and regional representation.  

Preventing new risks of exclusion 

The conversation highlighted how advanced technologies do not necessarily create new risks of exclusion. 

Hegoda explained that inclusion by design should be a foundation of every digital transformation journey, not an afterthought. 

“When you are implementing digital transformation, don’t listen to technical consultants. Please go to your personas and understand what their pain points are.” 

Pereira highlighted the importance of pairing infrastructure with education, because “there’s no point building the most advanced AI if my mother does not know how to use it.”  

This requires investment in awareness, training, and capacity-building to ensure that digital transformation benefits everyone, not just the digitally fluent. 

Lim cited a real-world example from Singapore where elderly citizens who did not know how to use a smartphone could still opt to use printed vouchers with QR codes printed on them.  

She called it the “0.5 version of digital infrastructure”. 

These hybrid approaches may not be perfect, but they are essential for bridging the gap between ambition and reality.  

“I think that as governments, we need to balance out that need for everything digital, everything on the phone with people who are digitally literate,” she noted. 

Click here to watch the full session.