EST-001

Bürokratt — Estonia's AI-Enabled Virtual Assistant Network for Public Services

Download PDF
Estonia Europe & Central Asia High income Operational Deployment (Limited Rollout) Likely

Information System Authority (RIA), under the Ministry of Economic Affairs and Communications

At a Glance

What it does LLMs for content creation, transformation and modality conversion — User communication and interaction
Who runs it Information System Authority (RIA), under the Ministry of Economic Affairs and Communications
Programme KrattAI / Bürokratt (public-sector virtual assistant network)
Confidence Likely
Deployment Status Operational Deployment (Limited Rollout)
Key Risks Model-related risks
Key Outcomes Reported improved service accessibility, 24/7 availability, reduced query burden for front-line staff.
Source Quality 7 sources — Government website / press release, Report (multilateral / development partner), Report (government / official)

Bürokratt is an AI-enabled virtual assistant network developed by the Estonian government to provide citizens with a unified, channel-agnostic interface for accessing public services, including areas relevant to social insurance and social protection. The initiative is led by the Information System Authority (RIA) under the Ministry of Economic Affairs and Communications, and forms a central component of Estonia's broader national AI strategy known as KrattAI, originally adopted by the government in July 2019. The name 'Kratt' draws from Estonian folklore, where a Kratt is a magical servant built from household items — a metaphor the government uses for AI assistants that handle routine administrative tasks on behalf of citizens.

The core technical architecture of Bürokratt is an interoperable network of AI-enabled chatbots deployed across the websites of public authorities. From the citizen's perspective, the system functions as a single communication channel through which they can access information and transact with government agencies. Currently, Bürokratt operates as a text-based chatbot on public agency websites, with voice-based interaction and sign language support planned for future phases. The system leverages large language models (LLMs) with retrieval-augmented generation (RAG) capabilities to understand natural language input and retrieve accurate, referenced responses from agency knowledge bases. The architecture uses Distributed Message Rooms (DMR), developed in collaboration between RIA and Microsoft engineers, which enable secure, encrypted inter-institutional communication so that queries can be routed between different agency chatbots within the Bürokratt network.

As of 2025, six public sector agencies have implemented Bürokratt, with over 30 additional state agencies expressing interest in adoption. The system is designed to eventually provide access to all approximately 3,000 Estonian government e-services, plus potentially private sector services, via both text and voice input. The platform's deployment roadmap for 2025 includes carrying out the deployment of an LLM/RAG system, developing a general knowledge module, and creating a central 'global classifier' component through which different Bürokratt instances can communicate securely with each other. Starting in 2026, Bürokratt is planned to move toward a model where each institution or domain operates its own personalised AI agent as part of a unified, cooperative network of agents, with a focus on building an LLM adapted specifically to the Estonian language.

In the social protection context, the retained sources support treating Bürokratt as a cross-government access layer that is relevant to benefits, pensions, family benefits, and other citizen-facing welfare services, rather than as a fully documented social-protection-specific deployment inside a single agency workflow. The proactive service model is designed so that government-initiated notifications can trigger service sessions — for example, when a child is born, the system can automatically notify parents and guide them through applying for child benefits, including specifying the bank account for payment. This aligns with Estonia's broader digital government philosophy of anticipatory, event-driven service delivery rather than requiring citizens to search for and initiate services themselves. However, specific deployment of Bürokratt within social insurance agencies such as the Social Insurance Board (Sotsiaalkindlustusamet) for pension or benefit-specific workflows has not been publicly documented in detail.

The entire Bürokratt codebase is released as open source under the MIT Licence, with code repositories available on GitHub. The Estonian government has explicitly made the platform available for reuse by other countries and private sector organisations. The open-source architecture is built on Estonia's X-Road interoperability infrastructure, which underpins the country's broader e-governance ecosystem. RIA has also collaborated with Microsoft to develop cloud-compatible prototypes running on Azure services, though the primary deployment remains on Estonian government cloud infrastructure under domestic jurisdiction and RIA oversight.

Bürokratt operates under Estonia's national data protection framework, which implements the EU General Data Protection Regulation (GDPR). The Data Protection Inspectorate (AKI) oversees how AI systems including Bürokratt process personal data. The system is designed to offer personalised services based on individual user data rather than generic information queries, which raises data protection considerations around the scope and sensitivity of personal data accessed during service sessions. Security measures include encrypted communication via the DMR architecture, logging, and integration with Estonia's national digital identity (eID) infrastructure for authentication. Human oversight is maintained through a model where the chatbot provides informational and advisory support rather than making binding administrative decisions — caseworkers review and validate outcomes where applicable, consistent with the system's advisory rather than decisional role.

The KrattAI strategy under which Bürokratt was developed committed the Estonian government to investing at least EUR 10 million in 2019-2021 for the implementation of approximately 50 AI use cases across government. Bürokratt represents the most visible and ambitious of these use cases. The system has reported improved service accessibility and 24/7 availability, with reduced query burden for front-line staff, though quantitative impact metrics have not been publicly detailed.

Classifications follow the DCI AI Hub Taxonomy. Hover over field labels for definitions.

Social Protection Functions

Implementation/delivery chain
Outreach/communications/sensitisation primaryProvision of payments/services
SP Pillar (Primary) The social protection branch: social assistance, social insurance, or labour market programmes. Social insurance
SP Pillar (Secondary) The social protection branch: social assistance, social insurance, or labour market programmes. Social assistance
Programme Name KrattAI / Bürokratt (public-sector virtual assistant network)
Programme Type The type of social protection programme, classified under social assistance, social insurance, or labour market programmes. View in glossary Other
System Level Where in the social protection system the AI is applied: policy level, programme design, or implementation/delivery chain. View in glossary Implementation/delivery chain
Programme Description Bürokratt is an interoperable network of AI chatbots across Estonian government agency websites, providing citizens unified access to approximately 3,000 public e-services, including some benefits- and pension-related public services. Six agencies live as of 2025, with 30+ additional agencies interested.
Implementation Type How the AI output is produced: Classical ML, Deep learning, Foundation model, or Hybrid. Affects validation, compute requirements, and governance profile. View in glossary Foundation model
Lifecycle Stage Current stage in the AI lifecycle, from problem identification through to monitoring, maintenance and decommissioning. View in glossary Integration and Deployment
Model Provenance Origin of the AI model: developed in-house, adapted from open-source, commercial/proprietary, or accessed via third-party API. View in glossary Adapted from open-source
Compute Environment Where the AI system runs: on-premise, government cloud, commercial cloud, or edge/device. View in glossary Sovereign or hybrid cloud
Sovereignty Quadrant Classification of data and compute sovereignty: I (Sovereign), II (Federated/Hybrid), III (Cloud with safeguards), or IV (Shared Innovation Zone). View in glossary II — Federated/Hybrid Governance
Data Residency Where the data used by the AI system is stored: domestic, regional, or international. View in glossary Domestic
Cross-Border Transfer Whether data crosses national borders, and if so, whether documented safeguards are in place. View in glossary Not documented
Is Agentic Whether the system autonomously plans and executes multi-step workflows, selecting tools and chaining actions with limited human intervention. View in glossary Partial
Agentic Pipeline Description of the chained workflow steps in the agentic pipeline. Chatbot receives user query, global classifier routes to appropriate agency Bürokratt instance, LLM/RAG retrieves from agency knowledge base, response returned. Multi-agent network planned for 2026.
Agentic Autonomy Degree of autonomy: fully autonomous, semi-autonomous (human checkpoints), or supervised (human approval at each step). Supervised
Override Points Where in the pipeline human review or override is triggered. Caseworkers review and validate outcomes; chatbot provides advisory/informational support only, not binding decisions
Decision Criticality The rights impact of the decision the AI supports. High criticality requires HITL oversight; moderate requires HOTL; low may operate HOOTL. View in glossary Low
Human Oversight Type Level of human involvement: Human-in-the-Loop (active review), Human-on-the-Loop (monitoring), or Human-out-of-the-Loop (periodic audit). View in glossary HITL
Development Process Whether the AI system was developed fully in-house, through a mix of in-house and third-party, or fully by an external provider. View in glossary Mix of in-house and third-party
Highest Risk Category The most significant structural risk source identified: data, model, operational, governance, or market/sovereignty risks. View in glossary Model-related risks
Risk Assessment Status Whether a formal risk assessment, informal assessment, or independent audit has been conducted for this system. Not assessed

Risk Dimensions

Data-related risks
Governance and institutional oversight risks
Operational and system integration risks

Impact Dimensions

Autonomy, human dignity and due process
Systemic and societal
  • Data minimisation controls
  • Human oversight protocol
CategorySensitivityCross-System LinkageAvailabilityKey Constraints
Unstructured and text-based contentNon-personalSingle source (no linkage)Currently available and usedPublic-service knowledge bases covering benefits, pensions, social insurance, health, and tax information used for RAG retrieval. Agency-specific content must be curated and maintained.

European Commission (2025). Government Virtual Assistant Bürokratt (OSOR case). Brussels: Interoperable Europe Portal. Available at: https://interoperable-europe.ec.europa.eu/collection/open-source-observatory-osor/government-virtual-assistant-burokratt (Accessed 31 Oct 2025).

View source Government website / press release

Information System Authority (RIA) (2025). Bürokratt. Tallinn: Information System Authority. Available at: https://www.kratid.ee/en/burokratt (Accessed 31 Oct 2025).

View source Government website / press release

RIA (2025). The vision of Bürokratt. Tallinn: Information System Authority. Available at: https://www.ria.ee/en/state-information-system/personal-services/burokratt (Accessed 23 Mar 2026).

View source Government website / press release

Interoperable Europe Portal (2023). Digital public services based on open source: case study on Bürokratt. Brussels: European Commission. Available at: https://interoperable-europe.ec.europa.eu/collection/open-source-observatory-osor/document/digital-public-services-based-open-source-case-study-burokratt (Accessed 23 Mar 2026).

View source Report (multilateral / development partner)

Microsoft (2024). RIA innovates its Bürokratt solution, in collaboration with Microsoft, to run on cloud services like Azure. Microsoft Customer Stories. Available at: https://www.microsoft.com/en/customers/story/1689174937531212442-information-system-authority-government-azure-en-estonia (Accessed 23 Mar 2026).

View source Report (government / official)

e-Estonia (2019). New e-Estonia factsheet: National AI 'Kratt' Strategy. Tallinn: e-Estonia Briefing Centre. Available at: https://e-estonia.com/new-e-estonia-factsheet-national-ai-kratt-strategy/ (Accessed 23 Mar 2026).

View source Government website / press release

e-Estonia (2022). Estonia's new virtual assistant aims to rewrite the way people interact with public services. Tallinn: e-Estonia. Available at: https://e-estonia.com/estonias-new-virtual-assistant-aims-to-rewrite-the-way-people-interact-with-public-services/ (Accessed 23 Mar 2026).

View source Government website / press release
Deployment Status How far the system has progressed into real-world operational use, from concept/exploration through to scaled and institutionalised. View in glossary Operational Deployment (Limited Rollout)
Year Initiated The year the AI system was first initiated or development began. 2022
Scale / Coverage The scale and geographic or population coverage of the deployment. Six public sector agencies live as of 2025; 30+ agencies interested; intended to cover all ~3,000 Estonian government e-services
Funding Source The source(s) of funding for the AI system development and deployment. Estonian government budget; KrattAI strategy allocated at least EUR 10 million for 2019-2021 across ~50 AI use cases
Technical Partners External technology vendors, academic partners, or development partners involved. Microsoft (DMR prototypes, Azure collaboration); RIA (lead development); open-source community contributors
Outcomes / Results Reported improved service accessibility, 24/7 availability, reduced query burden for front-line staff. Six agencies operational with positive response. Quantitative impact metrics not publicly detailed.
Challenges Specific deployment within social insurance and pension benefit workflows remains unverified in public sources. Voice and sign language interfaces not yet operational. LLM adaptation to Estonian language still under development. Scaling from 6 to 30+ agencies requires significant integration work.

How to Cite

DCI AI Hub (2026). 'Bürokratt — Estonia's AI-Enabled Virtual Assistant Network for Public Services', AI Hub AI Tracker, case EST-001. Digital Convergence Initiative. Available at: https://socialprotectionai.org/use-case/EST-001 [Accessed: 1 April 2026].

Change History

Updated 1 Apr 2026, 08:11
by system (system)
Created 30 Mar 2026, 08:39
by v2-import (import)