Podcast 🎧 & blog: Digital public infrastructure and the foundations of AI in government
Written by Federico Plantera, Researcher on tech policy and AI
Governments are racing to adopt AI in public services. EU-funded projects show the pace is only accelerating. But this push raises a deeper question: what lies beneath? Too often, the answer is the same – weak or uneven digital foundations.
Digital public infrastructure (DPI) can help integrate and connect siloed systems into a coherent digital government platform. So, where does it stand globally, as an enabler of AI development? What makes it work? And why does it matter more than ever for AI adoption? We explored these questions on this episode of the Digital Government Podcast with Krisstina Rao, now at Co-Develop, and outgoing Research Fellow at the Institute for Innovation and Public Purpose at University College London (UCL), where she led the work on the Digital Public Infrastructure Map.
First, the term itself. “Digital public infrastructure” may be new to policy debates, but the systems are not. Shared layers for identity, payments, and data exchange have existed for years. They often went by other names and long predate today’s hype. What’s less understood is just how widespread these systems already are.
More than a new term
“The definition changes based on who you ask, and this really matters,” Rao says at the outset. Search for digital public infrastructure online, and you will find a range of meanings – technical players emphasise interoperability, policy actors focus on shared systems, development practitioners on inclusion. Rao’s own emphasis, shaped by her social science background, falls on what these systems make possible: “Its focus on public sector capabilities and the ability of governments to facilitate participation in society and in markets.”
What makes this more than a rebranding of digital government? Rao traces a historical arc.
First came the era of departments digitising in silos – each building its own authentication layer and databases. Then came government service units that introduced common principles, but still focused on what the state itself delivered. Digital public infrastructure represents a further step: “The government is reframing how they think of service delivery, and of creating an ecosystem that is digitally enabled. And that ecosystem means society and markets.” The state shapes some of the conditions under which others – citizens, businesses, startups – participate.
This redefines what “public” means in the context. Rao points to the work of David Eaves and colleagues at UCL IIPP, who have examined what is actually public about digital public infrastructure. The answer turns out to have little to do with ownership. “The government doesn’t have to own everything. It’s actually very inefficient for the government to take on so many burdens. But they’re trying to create models where they’re enabling private participation – acting almost like a moulder of markets, rather than being passive in a system where it procures tech for public service delivery, and that’s where it draws the line,” Rao explains.
Practical example: “If a market exists and someone needs to participate in it – say, using digital infrastructure for a health startup, and you want to use an ID authentication platform to figure out who to deliver this service to – you should be able to, and the government should create this public utility.”
A utility that, within digital public infrastructure, looks different everywhere. Structurally, “not everyone is starting from the same stage of digital transformation or public sector capabilities when they begin thinking about what it means to share digital infrastructures within government or across governments,” Rao points out. Estonia built from a blank slate after 1991. Other countries digitised the analogue registries sector by sector. Countries in the Global South face yet other starting conditions. As she puts it, “The principle to take away is that digital public infrastructure is not a monolith.”
The Digital Public Infrastructure Map, which has researched systems across over 200 countries, confirmed what had been only a hypothesis. “Even within the ‘discourse bubble’ on the topic, we could only count five examples on our hands. But we realised there are stories that predate the term. Countries were doing it much earlier – they just weren’t calling it digital public infrastructure.” Having data to back this up, Rao says, “was a very validating experience, because previously it was only speculation.”
Is our data good enough?
Before working on the Map, Rao worked with a federal ministry in India to develop AI use cases for the health sector. The starting point turned out to have nothing to do with models. “Is our data good enough for a model to be built on top of it? Is it stored in PDFs? Is it an analogue? Are ‘apples actually apples, and oranges actually oranges’, in this data set? Because if you’re not putting those codes in, the AI is going to misread it.”
Digital public infrastructure and AI converge on the same layer, among others: data. Both depend on the same layer of the public sector technology stack – registries, identity systems, exchange protocols on one side; training and operational data on the other. If that layer is poor, both fail. And Rao spells out what a functioning digital public infrastructure setting actually does: “If you create an ID system, there’s a way to authenticate people. If you create a payment layer, there’s a way to enable individuals and actors to transact with each other. If you make sure this system is well-oiled through a data exchange system, then it’s interoperable. There are more use cases that can be built on top of it.”
Taking digital public infrastructure seriously, then, means treating it as the infrastructural approach to building the conditions under which AI in public services can work. “If you’re a government starting from scratch and you don’t have this layer already, investing in it now will help you create better data down the line for more use cases to be enabled on top of it,” Rao explains.
The relationship also runs the other way. AI creates demand for digital public infrastructure. The use cases that AI makes visible give government departments a reason to invest in the back-end systems they have been neglecting. Case in point, “If you know that down the line you need better quality data for the tax system in order to enable better identification of tax evasion, you’re going to invest in more interoperable layers and better data systems at the bottom.”
Investing in shared identity, data exchange, and transaction layers builds the foundation for AI adoption – and AI adoption, in turn, surfaces the gaps that make the case for those layers. “Investing in one kind of means you’re automatically creating fertile ground to invest in the other.”
Already in the EU, but under another name
This kind of infrastructure is often associated with India, Brazil, or Estonia. But Europe has been building shared digital layers for a long time – and the Digital Public Infrastructure Map reflects this. The EU’s TIPS system, enabling real-time transactions across European countries, is already in the dataset. “The EU used to call it maybe ‘shared infrastructure’, maybe ‘public digital infrastructure’. We are now just adding data to what the EU used to not call digital public infrastructure.”
The European integration agenda – with its emphasis on interoperability, single-market infrastructure, and cross-border service delivery – has created conditions for exactly these kinds of shared systems. But as Rao observes, the next step remains open. “The question now is how Europe works with markets to really leverage their ability to innovate, and make this layer more accessible to industry, to individuals, to different actors in society.”
What remains to be seen is whether European governments – at all levels – will connect the dots: recognising that these existing foundations, and the further investment they require, are potentially the same infrastructure on which their AI ambitions depend.
This podcast and podcast blog were produced within the project “EU Commission Project (24ES06/24DE33): Supporting regional entrepreneurship through the adoption of innovative technologies, including AI, in public services” with the financial assistance of the European Union via the Technical Support Instrument and implemented by the e-Governance Academy, in cooperation with the European Commission. The views expressed by the speakers in the project video are their own and do not necessarily reflect the official opinion of the European Union.
Interested in more?