In our recent report, we show that the most efficient governments are those that design their digital systems in line with democratic principles. The evidence is consistent across sectors and across countries
In our report, Government Efficiency in the Age of AI we argue that when participation, transparency, competition, subsidiarity, accountability and inclusion are built into the architecture, governments become more capable and more trusted. When they are not, efficiency gains flatten quickly, and technology begins to centralise power rather than distribute it.
This is the real challenge governments face with artificial intelligence (AI). The question is not how fast new tools can be deployed. It is whether the underlying digital structure can support democratic governance at scale. AI reinforces whatever sits beneath it. Good architecture accelerates reform. Poor architecture magnifies fragmentation and risk.
Below is a practical three-step path for building democratic digital public infrastructure, drawing from research and real-world experience.
1. Build a secure, trusted foundation
Most governments want to talk about AI, but the real leverage still sits in the architectural layer beneath it: base registries, unique identifiers, data exchange, digital identity and transparency mechanisms.
These components form the constitutional layer of the digital state. If they are mature, federated and governed through checks and balances, almost everything placed on top of them becomes safer, faster and more efficient. If they are fragmented or over-centralised, no amount of AI will compensate.
Three principles matter most:
First, it is federated. No single authority controls all core registries or data flows. Population, business, property and other registries are managed by separate institutions with clear legal boundaries. This reduces concentration of power and prevents mission creep.
Second, it is interoperable. Systems can share data legally, securely and in real time, allowing the once-only principle function in practice. It requires high-quality base registries, common identifiers and reliable data exchange. When implemented properly, it reduces costs, eliminates repeated paperwork and increases trust.
Third, it is transparent. Citizens must be able to see when their data is accessed, by whom and for what purpose. Tools such as personal data trackers provide visibility that reinforces consent and strengthens the social contract. Trust grows when people can verify how government uses their information.
These architectural decisions determine how safely and effectively a government can adopt AI.
2. Increase intelligence in systems and services
Once the foundation is in place, governments can focus on the core information systems that run major policy areas. Justice, health, tax, welfare, land use and emergency response all rely on how well data moves between institutions.
High-performing digital governments turn these domains into integrated super-domains. Within each one, data flows lawfully and in real time across the chain of actors. This reduces delays and improves accuracy. It also produces better evidence for policymaking.
To reach this level, governments need more than data sharing. They need shared standards and semantic alignment. When definitions vary between agencies, automation becomes unreliable and AI becomes risky. Common taxonomies, code
lists and data models are essential for scale.
These foundations support a shift in service delivery. Traditional online forms evolve into proactive services that anticipate needs. As standards mature, governments can move toward self-organising, agent-supported services. Instead of designing every cross-agency journey in advance, government exposes secure capabilities as APIs. Intelligent services assemble what the citizen needs in real time.
This approach reduces friction and avoids governance bottlenecks. It also makes AI safer. AI is effective when it is built on consistent structures, high- quality data and coherent processes. It exposes weaknesses when those conditions are not present.
3. Prepare democratic processes for real-time information
As governments rely more on real-time data and AI, the pace of administration accelerates. Democratic oversight does not accelerate automatically. Without adjustment, decisions can drift toward systems that operate faster than the mechanisms meant to oversee them.
Deterministic automation should remain the default for rule-based processes. It encodes legislation directly and preserves transparency and consistency. AI should be used where pattern recognition or unstructured data demand it, but always with human responsibility for decisions that affect rights or entitlements.
Accountability must be engineered. Models, data sources and decision logic should be documented, logged and auditable. Citizens must have the ability to question and appeal decisions influenced by automated systems.
Democratic participation must adapt as well. Secure digital identity and trusted digital channels can support more continuous consultation and more frequent engagement. The objective is not speed. It is legitimacy at the pace of real-time information.
The long-term risk is not rapid digitalisation but unexamined drift. Governments that strengthen transparency, oversight and participation early will retain democratic resilience as AI becomes more pervasive.
The structural choices that shape digital government
Digital infrastructure is often treated as a technical project. In practice, it is institutional design. Registries, identity, interoperability and automation determine how government functions and how power is distributed.
If democratic principles guide these choices, governments become more efficient and more trusted. If they are treated as secondary concerns, no amount of AI will correct the imbalance later.
The opportunity is clear: build digital systems that scale trust, not only transactions. The decisions governments make now will define the kind of digital state they build.
These themes, along with a practical roadmap for policymakers, are explored in detail in our report Government efficiency in the age of AI: Toward resilient and efficient digital democracies. This paper is co-authored by Dr Mihkel Solvak (Tartu University), Dr Ott Velsberg (Estonian Ministry of Justice and Digital Affairs) and Dr Keegan McBride (Tony Blair Institute for Global Change), with review contributions from Dr David Ronfeldt.
This work is licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International.
