An AI-powered platform that helps civil servants navigate complex, cross-cutting policy challenges — turning weeks of fragmented analysis into hours of structured insight.
Today's policy challenges — demographic decline, climate adaptation, energy security, post-war economic pressures — don't respect ministry boundaries. Yet governments remain structured in silos with linear planning cycles designed for simpler times. Civil servants face compounding crises with fragmented data, disconnected expertise, and rigid processes. The gap between the complexity of problems and the capacity to respond is growing every year.
Latvia exemplifies this challenge — and offers the ideal proving ground for a solution. A country of 1.86 million navigating simultaneous pressures across 14 ministries, with a compact civil service that makes whole-of-government transformation possible within a single project cycle.
Latvia's size isn't a limitation — it's what makes this project possible. Whole-of-government deployment across 14 ministries within 36 months is something you simply can't attempt first in larger EU states. If CrossGov AI works in Latvia, it becomes a replicable model for the 20+ EU member states facing identical pressures.
Ranked 6th in EU digital public services. Government SSO, digital identity, and open data APIs are already in place — our tool integrates into existing infrastructure, not alongside it.
VAS (Latvian School of Public Administration) trains every ministry's civil servants. This gives us a built-in distribution channel that no single ministry could provide.
92% of participants in our Humble Governance learning journey rated the methodology as directly applicable. Latvian civil servants are actively asking for tools like this.
Demos Helsinki has active governance partnerships in Finland, Estonia, and across EU institutions. The jump from proven Latvia deployment to multi-country adaptation is already mapped.
CrossGov AI is a decision intelligence platform built on Google Gemini that transforms how civil servants approach complex policy challenges. Instead of weeks of manual research and ad hoc consultation, officials get structured AI support across the full decision workflow — from framing the problem to testing solutions.
Starting from a user-described challenge, AI surfaces connections to adjacent policy domains, identifies relevant datasets, and generates interactive maps revealing dependencies that siloed analysis misses.
Rapidly aggregates local government data, international policy precedents, and stakeholder perspectives into actionable briefings — in hours instead of the weeks manual research requires.
An agentic AI that guides civil servants through designing safe-to-fail policy experiments — defining hypotheses, identifying metrics, and anticipating consequences before committing to reform.
CrossGov AI augments strategic decision-making — it never makes decisions autonomously. Civil servants review, validate, and act on every output. The AI surfaces what humans would miss; humans exercise the judgment AI cannot.
A civil servant describes a policy challenge in natural language. The AI identifies cross-domain connections and generates an interactive problem map.
The system gathers and synthesizes evidence from local government data, EU repositories, OECD databases, and academic research into a structured briefing.
Instead of committing to large-scale reform, the agentic AI helps design small-scale policy experiments with clear hypotheses and success metrics.
Results feed back into the knowledge base, building institutional memory that persists beyond individual civil servant tenures.
CrossGov AI doesn't start from scratch. It operationalizes the Humble Governance methodology — developed with Columbia Law School's Prof. Charles Sabel, validated through the Finnish Prime Minister's Office, and winner of the 2021 Creative Bureaucracy Festival Award.
Measured through pre/post survey data during our 10-week Humble Governance learning journey with senior Latvian civil servants (January–March 2026).
Participants across multiple ministries confirmed the methodology maps to their actual workflow challenges — generative problem mapping and evidence synthesis were identified as the highest-value applications.
The structured, safe-to-fail approach gave civil servants confidence to propose small-scale experiments — exactly the behavior CrossGov AI's third module is designed to enable at scale.
Tested via a 3-day symposium and 10-week learning journey delivered through VAS, generating baseline data and firsthand insight into workflows, pain points, and readiness that shapes our AI design.
Every technology choice is driven by two principles: leverage Google's ecosystem for AI capabilities, and ensure full compatibility with EU data governance (GDPR, EU AI Act).
Primary LLM via Vertex AI for generative problem mapping, evidence synthesis, and natural language interaction in English and Latvian.
Orchestration framework for agentic workflows — multi-step evidence gathering, experiment design guidance, and adaptive follow-up queries.
Vector databases (Pinecone/Weaviate) connecting Gemini to curated knowledge bases: Latvia's open data portal, OECD policy databases, EUR-Lex, and internal government documents.
Hosting, authentication, and compliance infrastructure. All data stays within EU regions (Finland/Netherlands). Government SSO integration via Google Cloud Identity.
Frontend integrated into existing government digital infrastructure. Bilingual interface (Latvian/English) with accessible, government-standard UX.
Full codebase published under Apache 2.0 license by Month 24, enabling any government to adapt and deploy. Contributing back to the AI-for-governance commons.
The partnership structure is designed for adoption, not just development. Demos Helsinki brings the methodology, AI expertise, and international network. VAS brings the government access, training infrastructure, and institutional legitimacy that turns a tool into a standard practice.
Independent think tank founded in 2005. Worked with governments in 30+ countries. €6.1M revenue, 74 staff. Creator of the Humble Governance framework (Creative Bureaucracy Festival Award 2021, developed with Columbia Law School's Prof. Charles Sabel). Brings: governance methodology, AI solution design, international scaling network, Google accelerator relationship.
Latvia's official civil servant training institution with reach across all 14 ministries. Active partner since October 2025, building on Demos Helsinki's collaboration with Latvia's State Chancellery since 2020. Director Agita Kalviņa has confirmed institutional support. Brings: cross-government access for co-design and testing, integration into official training curricula, policy domain expertise, and the institutional pathway no single ministry could provide.
Latvia is the testbed, not the ceiling. By Month 24, the platform, methodology integration, and deployment playbook will be published as open source — enabling any government to adapt it.