Estonia eyes cross-border interoperability for Bürokratt, its ‘Siri of public services’

By Si Ying Thian

The government’s Chief Data Officer Dr Ott Velsberg shares that an open and modular systems approach, even for a region as diverse and multilingual nations as APAC, is key to creating a network of interoperable virtual assistants for government services.

Virtual assistants are commonplace in government services today. To improve the citizen experience, the focus must shift from simply answering inquiries to redesigning these assistants to navigate bureaucracy, says Dr Velsberg. Image: e-Estonia

For most people, the word bureaucracy is synonymous with endless forms and frustrating delays. 

 

The Estonian government wants to overhaul the bureaucratic experience for its citizens by building an entire interoperable network of artificial intelligence (AI) systems, not just one government virtual assistant, to simplify access to various services. 

 

By using one device in one communication session, Bürokratt (a play on the word “bureaucrat”) allows a person to use voice or text commands to get everything they need without having to navigate multiple interfaces. 

 

Bürokratt is a network of chatbots on Estonian government agency’s websites that enables people to obtain information and services via virtual assistants. 

 

The government has made it clear that while it can’t rival Big Tech’s virtual assistants, it can position its role in enabling unlimited scale of integration within and between the public sector, private sector and globally.  

 

Speaking to GovInsider, the government's Chief Data Officer Dr Ott Velsberg says to date, Bürokratt has integrated services across 18 organisations with more to come. 

Dr Ott Velsberg is the Chief Data Officer of the Estonian Government. Image: Dr Velsberg
 

He added that the government is now focusing on establishing cross-border interoperability for Bürokratt so that virtual assistants deployed in different countries can seamlessly connect and collaborate with each other. 

 

On Bürokratt’s website, the government underlined its plan to establish the foundational infrastructure for the virtual assistant network within this year. 

 

These steps include deploying a large language model (LLM) and retrieval augmented generation (RAG) that would be able to pull accurate, up-to-date answers from specific databases; develop a shared knowledge base for the entire network; and create a global classifier to securely manage communication and ensure smooth handovers between individual assistants from different agencies. 

 

Starting 2026, the government will focus on personalised AI agents and widening its interoperable network, including building an Estonian-adapted LLM and making public sector services more accessible and human-centred. 

 

To subscribe to the GovInsider bulletin, click here

Lessons for APAC 

 

Today, virtual assistants are commonplace in government services. But to truly improve the citizen experience, the focus must shift from simply answering inquiries to redesigning these assistants to navigate bureaucracy, says Dr Velsberg. 

 

He adds that the modular design of Bürokratt has allowed governments to start small and scale fast.  

 

For example, governments can start with simple notifications, then expand to proactive benefits and ultimately, to fully event-driven services to align with their readiness, he says. 

 

What a diverse and multilingual region like Asia-Pacific can learn from Estonia’s experience with Bürokratt lies in its foundation principles, not its specific technology, he explains. 

 

These principles include open interfaces, built-in identity and consent, local-language models, and strong guardrails for trust. 

 

Dr Velsberg adds that Bürokratt also redefined the relationship between the state and its citizens, from “you come to government” to “government accompanies you wherever you are.” 

 

The government’s official blog highlighted plans to integrate Bürokratt with Finland’s AuroraAI, which would allow pregnant Estonian woman in Finland, for instance, to use Bürokratt to seamlessly arrange hospital transport and manage all her related needs, regardless of her being abroad.  

CDO as a cross-functional orchestrator 

 

In the age of AI, Dr Velsberg role has evolved to that of a “cross-functional orchestrator,” he says, bridging data, AI, security and service design. 

 

Redesigning services and processes is more important than simply stacking new solutions or models on top of existing ones, he says. 

 

His team has supported Estonia's entire journey - from e-government through digital government to the post-digital era - by identifying and solving problems, integrating solutions, and focusing on maximising impact and making public administration more efficient and effective for both citizens and businesses. 

 

According to Dr Velsberg, the government is currently in the post-digital era, where it is possible for services to have no analogue in the non-digital world. 

 

“Here, the focus is on creating personalised, proactive services that leverage the opportunities of the data economy and AI,” he explains. 

 

Since Estonia rolled out its first AI strategy in 2019, he highlights that there has been a growing emphasis on data governance and data-driven decision-making. 

 

To support the state's ability to provide personalised and proactive services, Dr Velsberg and his team are focusing on improving data accessibility, ensuring sufficient talent and skills, strengthening cybersecurity and data protection by design, as well as securing sovereign compute for sensitive workloads. 

Responsible AI and proactive data governance by design

 

Since 2019, the government has embedded a number of responsible AI requirements into all funded projects, says Dr Velsberg.

 

Any project receiving support must carry out a data protection impact assessment, address cybersecurity risks, conduct data quality checks, implement the data tracker to ensure transparent processing and abide to the principle of purpose limitation.

 

He adds that the government is currently working to make algorithmic transparency mandatory for all AI developers.

 

This new rule will require organisations to provide a clear overview of the solution's function, intended purpose and how it is monitored once in use, he explains. The government will provide additional support to mitigate bias, assess privacy impact and explainability.

 

Back in 2019 shortly when he was appointed as CDO, Dr Velsberg told GovInsider that leveraging an open data economy depended on getting public literacy, data quality and governance right. 

 

“Open data remains a powerful growth engine [for innovation], but not all data is created equal,” he tells us today. He adds that citizen empowerment and transparency are key to enabling this growth engine.  

 

First, the government has strategically prioritise investment in certain high-value domains like spatial data, language data and satellite imagery to build national data capacity.  

 

It has established ESTHub as a dedicated satellite data portal and processing platform. Additionally, it is developing a language data space that allows individuals to decide who can process their data and under what conditions.

 

Secondly, there are also a few initiatives aimed at enhancing overall data transparency and citizen control. 

 

Central to this effort is the new government data portal launched in June, which offers a holistic view of all government data (both restricted and non-restricted) as well as its processing and access methods. 

 

This is accompanied by two other services including the data tracker, a centralised dashboard that allows citizens to see when, how and why their personal data has been accessed in national databases; as well as the consent service that allows citizens to voluntarily grant companies permission to access their personal data held in national databases. 

 

More than one million data consents have been granted, with the main use cases being in banking and healthcare, says Dr Velsberg. 

 

For other news about public private integration and open data, read more here: